I'm getting the urge to build a low energy server round a mini-ITX
board, a couple of 2.5" drives and a DVD writer. Low power consumption
would be nice because it will run 24/7. 

The board needs to at least match the performance of my current server, so
should be 1 - 1.5 GHz and with at least 0.5 GB RAM. It should support
at least two Hard drives, the aforementioned DVD writer, at
least one 10/100 Mbs ethernet port and at least two USB2.0 ports (4 would
be better). A parallel port would be nice too. 

I'll be running a current Fedora distribution on it. 

Are there any chipsets and/or makes of board that are to be
particularly avoided? Any obvious preferences? 
 
TIA

-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

In article <pan.2008.05.17.12.26.07.623271@see.sig.for.address>,
Martin Gregorie  <martin@see.sig.for.address> wrote:
>I'm getting the urge to build a low energy server round a mini-ITX
>board, a couple of 2.5" drives and a DVD writer. Low power consumption
>would be nice because it will run 24/7. 
>
>The board needs to at least match the performance of my current server, so
>should be 1 - 1.5 GHz and with at least 0.5 GB RAM. It should support
>at least two Hard drives, the aforementioned DVD writer, at
>least one 10/100 Mbs ethernet port and at least two USB2.0 ports (4 would
>be better). A parallel port would be nice too. 
>
>I'll be running a current Fedora distribution on it. 
>
>Are there any chipsets and/or makes of board that are to be
>particularly avoided? Any obvious preferences? 

I may be a bit biased, as I build Linux "appliances" out of Mini ITX
boards, but FWIW, here's my take on it...

A VIA CN1300 is 1.3GHz and can take up to 1GB of RAM (which, given
todays prices, you might as well do). It has a CPU fan. There is a 1GHz
version of the board without the fan if you want super quiet.

My favourite place: http://linitx.com/

Their web site is linked to their stock-control system unlike another
place I used to use, and they keep you informed of order progress, read
your emails in a timely manner and appear to be all-round good guys...

There are many other board solutions - some with faster processors and
sockets for Core2 duo chips too - but these will then push the power
consumption up. My 1GHz boards idle at about 15W with 512MB of RAM and a
128MB flash IDE drive.

See: http://unicorn.drogon.net/power.jpg

For a 1.5GHz board, have a look at:

  http://linitx.com/viewproduct.php?prodid=11470

This has a DVI connector too, and the full audio kit, IDE and SATA,
however its "legacy free", so no PS/2 mouse/keyboard or parallel
ports... There are headers for 4 more USB2 ports

The down-side of the really low-power ones is the CPU cache. These are
128KB only, and this might be the thing that slows you down. They don't
have the full compliment of MMX/SSE instructions either, so graphics
crunching might not be "optimal". Saying that, a friend built a couple
of these boxes for a local church youth group (1GHz processor, 1GB RAM
in a wall mount case so it could be screwed to the desk!) and he put XP
on them and is happily running Firefox and OpenOffice on them, so with a
Linux distro you ought to be OK for general day to day use, web browsing,
etc., but don't expect the earth when you go to very busy web sites!

Gordon

Martin,

I've been thinking about doing this for ages to replace my ageing web-facing
server, let us know how you get on.

Gordon,

thanks for the information, I've been browsing the mini-itx.com web-site and
hadn't seen the linitx one. It's very handy to get recommendations from
people who have actually used the service.

-- 
Geoff                                           Registered Linux user 196308
Replace bitbucket with geoff to mail me.

On Sat, 17 May 2008 13:26:08 +0100, Martin Gregorie wrote:

> I'm getting the urge to build a low energy server round a mini-ITX
> board, a couple of 2.5" drives and a DVD writer. Low power consumption
> would be nice because it will run 24/7. 
> 
> The board needs to at least match the performance of my current server, so
> should be 1 - 1.5 GHz and with at least 0.5 GB RAM. It should support
> at least two Hard drives, the aforementioned DVD writer, at
> least one 10/100 Mbs ethernet port and at least two USB2.0 ports (4 would
> be better). A parallel port would be nice too. 
> 
> I'll be running a current Fedora distribution on it. 
> 
> Are there any chipsets and/or makes of board that are to be
> particularly avoided? Any obvious preferences? 
> 

I have been running a low power mini-ITX based server for about 18 months
since my laptop based server packed up the ghost.

For the motherboard I use the EN12000G, it has a 1.2GHz fanless Via C7 and
1Gbps Ethernet. For the price of the RAM I would stick in 1GB, no point in
skimping on that front.

For hard drives a use a couple of "Always-On" rated Hitachi 2.5" 7200RPM
IDE drives in software RAID-1. I cannot stress how important it is to get
the "Always-On" versions of the drives. I got IDE, because the at the time
the SATA "Always-On" where not available. The EN12000G has two SATA and
two IDE, so you could use either. Let me stress this point again you
*REALLY* *REALLY* need to get "Always-On" rated drives. Normal laptop
drives give out after one to two years when on 24x7. Been through several
ordinary laptop drives running low power servers. Got fed up and
switched to the "Always-On" variant. Not seen a failure since.
www.span.com have 200GB 2.5"/7200RPM always on SATA available.

http://www.span.com/catalog/product_info.php?cPath=23_512_628&products_id=16005

An interesting alternative would be the new WD VelociRaptor drives, 300GB
at 10,000RPM taken out their headsink thing. Their power requirements are
supposed to be not that much worse than an laptop drive.

For power I use a picoPSU at 60W. The idea here is that with a 6-26V DC
input I can build a battery backed PSU rather than a UPS. The concept is
that the a 12V lead acid can provide the power when the mains goes away,
without conversion back to mains. That said I have not got around to
actually building it yet. The only issue is this version does not produce
much on the 12V line so all storage has to be laptop based.

http://www.mini-box.com/s.nl/sc.8/category.13/.f

There is a new pico M3-ATX designed for vehicle applications rated at 125W
that produces much more juice on the 12V line so you can use desktop
storage devices. Also has battery deep discharge prevention built in
making a battery backed PSU design simpler.

I used a Pack-Box A4 for the case, with the fan removed (my system is
fanless for increased reliability).

http://www.icp-epia.co.uk/index.php?act=viewProd&productId=321

One point with this is I had to do some metal work on the case. First the
hole for the 12V DC input needed drilling out to take the input jack on
the picoPSU. Then I had to drill some mounting holes for the two 2.5" hard
drives as none of the pre-drilled holes matched up!!

Under load the whole lot draws just over 30W at the plug, and at the
moment it is using a cheap and not very efficient SMPS to get the 12V.
Measuring the 12V line indicates it draws about 21W under load.

Mine is also fitted with a PCI ADSL card, cause it is neater and lower
powered than an external modem. A bit nasty as it requires binary kernel
modules. For the wireless, with a little difficulty I secured a suitable
Prism chipset based USB thumb drive thing from eBay, which provides the
wireless connectivity. Finally I have my own design radio clocks (both MSF
and DCF77) hooked up to a serial port. The whole lot is mounted on the
wall under the stairs.

I run with Debian Etch on the box, though given time again I would
probably go with CentOS 5, especially now with the Fedora EPEL 5 repo. The
idea that I would not need to do a base OS upgrade in the lifetime of the
hardware is quite attractive. I am also cursing that I need to repurchase
my GoDaddy SSL cert for the box...

You mention a parallel port, my suggestion is do yourself a favour and get
an external JetDirect on eBay, that way you can divorce the server from
the printer. You can usually pick up an EX, EXplus or 170 for under £10
delivered! Frankly a printer without a Ethernet port is not a printer.


JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan (at) buzzard.me.uk
St. Andrews, United Kingdom.

On Sat, 17 May 2008 15:59:35 +0100
Geoffrey Clements  wrote:

> Martin,
> 
> I've been thinking about doing this for ages to replace my ageing
> web-facing server, let us know how you get on.
> 
> Gordon,
> 
> thanks for the information, I've been browsing the mini-itx.com
> web-site and hadn't seen the linitx one. It's very handy to get
> recommendations from people who have actually used the service.
> 
I bought from linitx, and was seriously disappointed.

First, I specifically asked about a quiet box (yes, I phoned).
The case I got was bloody noisy.

Second, I had constant hardware trouble.  The onboard networking
died fairly quickly, so I had to put a network card in the
single PCI slot.  Then the PCI slot started playing sillybuggers.
Then the whole thing started powering down at random and without
warning, until it failed altogether.

I replaced it with a conventional Sempron-based box based on
cheapest-available components from DABS (together with the
disc and memory from the linitx, which are bog-standard
components and are fine).  It's quieter, and has yet to
give hardware trouble - apart from being disinclined to work
with some of my older PCI cards.

-- 
Nick Kew

Application Development with Apache - the Apache Modules Book
http://www.apachetutor.org/

On Sat, 17 May 2008 17:48:51 +0100, Jonathan Buzzard wrote:

> 
> For the motherboard I use the EN12000G, it has a 1.2GHz fanless Via C7 and
> 1Gbps Ethernet. For the price of the RAM I would stick in 1GB, no point in
> skimping on that front.
>
OK, I'll add C7 systems to my list. I was thinking in terms of 1GB RAM in
any case, though it still seems obscene when you consider that I used to
use a machine that supported around 200 green screens and a dozen
database systems with 16 MB RAM. My current server certainly went faster
when I moved it from 256 MB to 512 MB.


> For hard drives a use a couple of "Always-On" rated Hitachi 2.5" 7200RPM
> IDE drives in software RAID-1. I cannot stress how important it is to
> get the "Always-On" versions of the drives.
>
The trigger for this interest was being given a pair of Fujitsu 20 GB 2.5"
drives that came out of Pace HDD VCRs. Are these likely to be always on?
Is there any way I can easily find out?

> 
> For power I use a picoPSU at 60W. The idea here is that with a 6-26V DC
> input I can build a battery backed PSU rather than a UPS. The concept is
> that the a 12V lead acid can provide the power when the mains goes away,
> without conversion back to mains. That said I have not got around to
> actually building it yet. The only issue is this version does not
> produce much on the 12V line so all storage has to be laptop based.
>
Good idea. I like the idea of having the battery continuously on,
especially as it should make a decent buffer against mains
fluctuations and spikes. You'd really need a dual mode charger (one with
bulk charge and trickle modes) if it is to do much more than keep up with
the computer's demand. Otherwise it will just wreck the battery, but I
expect you know this already. Does anybody know if the Maplins dual mode
chargers can switch back to bulk charge from trickle mode  without being
switched off and on again? They're good chargers though not terribly
efficient judging by their normal operating temperature.

> I used a Pack-Box A4 for the case, with the fan removed (my system is
> fanless for increased reliability).
>
I like the look of the Morex Venus 669 case:
http://www.mini-itx.com/store/?c=3#venus
because I don't need the rig to be all that small and I like the
tidiness of an internal PSU.

> Finally I have my own design radio clocks (both
> MSF and DCF77) hooked up to a serial port. The whole lot is mounted on
> the wall under the stairs.
>
I forgot to mention serial ports. Thanks for the reminder: I need two  of
those too. One for an MSF  clock (your design!) and one for transferring
data (devices like a Garmin GPS II+ and an EW flight recorder).

> You mention a parallel port, my suggestion is do yourself a favour and
> get an external JetDirect on eBay, that way you can divorce the server
> from the printer.
>
Yes, you're dead right. I even have the network printer already: a
visit to Cambridge Resale a couple of years back got me a network-ready
HP LJ5, but so far laziness and a lack of spare ports on my CAT5 hub has
meant that it plugged into the server's parallel port, which also allows
it to be shared with my truly ancient OS-9/68K system via an equally
ancient auto switch.


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

On Sat, 17 May 2008 14:10:11 +0000, Gordon Henderson wrote:

I remembered your activities in this area and hoped you'd reply.

> For a 1.5GHz board, have a look at:
> 
>   http://linitx.com/viewproduct.php?prodid=11470
> 
> This has a DVI connector too, and the full audio kit, IDE and SATA,
> however its "legacy free", so no PS/2 mouse/keyboard or parallel
> ports... There are headers for 4 more USB2 ports
> 
> The down-side of the really low-power ones is the CPU cache. These are
> 128KB only, and this might be the thing that slows you down.
>
Thanks for the tip. Given a suitable case It looks as though I can simply
add a PCI serial card provided that I choose a board with a PCI socket: I
forgot to mention that I need a pair of those too (one for a JAB MSF time
receiver and the other for data transfer to GPS II+, EW logger and
Parallax STAMPs.

However, I notice that, while the board mentioned above has far more
multimedia capability that I'm likely to use, it's spec doesn't give the
cache size. Is this data that's usually given?


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

On 18/05/2008 00:34, Martin Gregorie wrote:

> for a JAB MSF time receiver 

You mean GBZ receiver :-P

> the board mentioned above has far more
> multimedia capability that I'm likely to use, it's spec doesn't give the
> cache size. Is this data that's usually given?

The cache is built-in to the C7 processor, so it fixed at 128Kb
http://www.via.com.tw/en/products/processors/c7/specs.jsp

On Sun, 18 May 2008 09:31:05 +0100, Andy Burns wrote:

> On 18/05/2008 00:34, Martin Gregorie wrote:
> 
>> for a JAB MSF time receiver 
> 
> You mean GBZ receiver :-P
> 
I've always heard Rugby referred to as an MSF60 time source. I'm using the
Galleon Rx with a board and software designed by Jonathan Buzzard - hence
the terminology.

> The cache is built-in to the C7 processor, so it fixed at 128Kb
> http://www.via.com.tw/en/products/processors/c7/specs.jsp
>
Thanks for the info. If the C7 performance is limited by cache size, is
there another chip with a better bang for the watt that I should be
looking at. If I can keep the mini-ITX board under 25 watts and the full
system below 40 (with DVD idle) so much the better.


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

On 18/05/2008 09:50, Martin Gregorie wrote:

> I've always heard Rugby referred to as an MSF60 time source. 

The callsign of the MSF transmitter changed to GBZ when it "moved" from 
Rugby to Anthorn, that's all.

> If the C7 performance is limited by cache size, is
> there another chip with a better bang for the watt that I should be
> looking at.

Most of the embedded CPUs have smaller cache, to save power.

> If I can keep the mini-ITX board under 25 watts and the full
> system below 40 (with DVD idle) so much the better.

I still keep an eye on small boards, but it's been many years since I've 
used them really.

In article <pan.2008.05.17.23.34.36.517343@see.sig.for.address>,
Martin Gregorie  <martin@see.sig.for.address> wrote:
>On Sat, 17 May 2008 14:10:11 +0000, Gordon Henderson wrote:
>
>I remembered your activities in this area and hoped you'd reply.
>
>> For a 1.5GHz board, have a look at:
>> 
>>   http://linitx.com/viewproduct.php?prodid=11470
>> 
>> This has a DVI connector too, and the full audio kit, IDE and SATA,
>> however its "legacy free", so no PS/2 mouse/keyboard or parallel
>> ports... There are headers for 4 more USB2 ports
>> 
>> The down-side of the really low-power ones is the CPU cache. These are
>> 128KB only, and this might be the thing that slows you down.
>>
>Thanks for the tip. Given a suitable case It looks as though I can simply
>add a PCI serial card provided that I choose a board with a PCI socket: I
>forgot to mention that I need a pair of those too (one for a JAB MSF time
>receiver and the other for data transfer to GPS II+, EW logger and
>Parallax STAMPs.

I suspect I'd look for USB serial/parallel devices these days... Not
sure if the JAB box will work over one though - maybe I'll have a go
with mine one day, as it's curently gathering dust...

>However, I notice that, while the board mentioned above has far more
>multimedia capability that I'm likely to use, it's spec doesn't give the
>cache size. Is this data that's usually given?

It's one of those things you need to dig deep to get the specs on. In my
case from the VIA website, and 1st hand experience:

This is one of my older dev. boxes: (About 5 years old now)

  processor       : 0
  vendor_id       : CentaurHauls
  cpu family      : 6
  model           : 7
  model name      : VIA Samuel 2
  stepping        : 3
  cpu MHz         : 533.377
  cache size      : 64 KB
  fdiv_bug        : no
  hlt_bug         : no
  f00f_bug        : no
  coma_bug        : no
  fpu             : yes
  fpu_exception   : yes
  cpuid level     : 1
  wp              : yes
  flags           : fpu de tsc msr cx8 mtrr pge mmx 3dnow
  bogomips        : 1067.57

This is a 1GHz C3:

  processor       : 0
  vendor_id       : CentaurHauls
  cpu family      : 6
  model           : 10
  model name      : VIA Esther processor 1000MHz
  stepping        : 9
  cpu MHz         : 997.560
  cache size      : 128 KB
  fdiv_bug        : no
  hlt_bug         : no
  f00f_bug        : no
  coma_bug        : no
  fpu             : yes
  fpu_exception   : yes
  cpuid level     : 1
  wp              : yes
  flags           : fpu vme de pse tsc msr pae mce sep mtrr pge cmov pat
  clflush acpi mmx fxsr sse sse2 tm pni est tm2 rng rng_en ace ace_en ace2
  ace2_en phe phe_en pmm pmm_en
  bogomips        : 1996.57

This is a 1.3GHz version of the above:

  processor       : 0
  vendor_id       : CentaurHauls
  cpu family      : 6
  model           : 10
  model name      : VIA Esther processor 1300MHz
  stepping        : 9
  cpu MHz         : 1296.846
  cache size      : 128 KB
  fdiv_bug        : no
  hlt_bug         : no
  f00f_bug        : no
  coma_bug        : no
  fpu             : yes
  fpu_exception   : yes
  cpuid level     : 1
  wp              : yes
  flags           : fpu vme de pse tsc msr pae mce sep mtrr pge cmov pat
  clflush acpi mmx fxsr sse sse2 tm pni est tm2 rng rng_en ace ace_en ace2
  ace2_en phe phe_en pmm pmm_en
  bogomips        : 2595.41

Also note that VIA have different processor specs too - even for the
same clock speed. This is a 1GHz board - chosen as it has 2 on-board
Ethernet sockets (it's in-use as a router)

  processor       : 0
  vendor_id       : CentaurHauls
  cpu family      : 6
  model           : 9
  model name      : VIA Nehemiah
  stepping        : 10
  cpu MHz         : 1000.076
  cache size      : 64 KB
  fdiv_bug        : no
  hlt_bug         : no
  f00f_bug        : no
  coma_bug        : no
  fpu             : yes
  fpu_exception   : yes
  cpuid level     : 1
  wp              : yes
  flags           : fpu vme de pse tsc msr cx8 apic sep mtrr pge cmov pat
  mmx fxsr sse rng rng_en ace ace_en
  bogomips        : 2001.82
  clflush size    : 32

Gordon

In article news:<pan.2008.05.17.12.26.07.623271@see.sig.for.address>, 
Martin Gregorie wrote:
> Are there any chipsets and/or makes of board that are to be
> particularly avoided?

As you're specifically after low power I'd recommend one of the VIA 
mini-ITX boards. There are other makes of board that use VIA chips/CPU, 
and they are probably as good, but I ave no direct experience of them.

Mini-ITX boards are available that use low-power version of the AMD and 
intel x86 CPUs -- an these are more powerful -- but they will be more 
energy hungry than the VIA boards.

As a very rough guide you should expect a VIA C3/C7 CPU to be about as 
fast as a Celeron of the same clock speed -- the VIA chips have less L2 
cache than a Celeron and the mini-ITX boards mostly don't support 
dual-channel RAM (there's only one slot) which slows them down.

Something like the VIA EN1200E might suit? It's 1.2GHz, fanless, and 
has 2xIDE, 2xSATA and 6 USB ports (4 on the back, and a header for 2 
more).

> I'll be running a current Fedora distribution on it.

I ran Fedora Core 1 (when it was current) on a VIA M10000 board without 
any problems. I don't anticipate that you'll have any with a more 
recent Fedora on a more recent VIA board. The only thing to watch for 
is that linux graphics drivers for the CN700 Northbridge are not 
available (or weren't, last time I looked) ... but it'll run in VESA 
mode, which will be good enough for the sort of server setup you seem 
to be considering.

Cheers,
 Daniel.

In article news:<pan.2008.05.17.22.53.14.13291@see.sig.for.address>, 
Martin Gregorie wrote:
> I like the look of the Morex Venus 669 case:
> http://www.mini-itx.com/store/?c=3#venus

I have the earlier version of the Venus case ... it's nicely put 
together but the fans are bloody noisy.

Cheers,
 Daniel.

On Sun, 18 May 2008 11:19:21 +0100, Daniel James wrote:

> Mini-ITX boards are available that use low-power version of the AMD and 
> intel x86 CPUs -- an these are more powerful -- but they will be more 
> energy hungry than the VIA boards.
>
Are you meaning Celerion and equivalent mainstream laptop CPUs here? If
so, how do their power consumption and performance compare with a C7? What
is there that's a bit nippier than the Celerion, but still fairly easy on
power? I'm not intending to run a Core Duo mini-ITX because the purpose of
building this box is to save power, not space.

> I ran Fedora Core 1 (when it was current) on a VIA M10000 board without
> any problems. I don't anticipate that you'll have any with a more recent
> Fedora on a more recent VIA board. 
>
That's good to know. Thanks.

> The only thing to watch for is that
> linux graphics drivers for the CN700 Northbridge are not available (or
> weren't, last time I looked) ... but it'll run in VESA mode, which will
> be good enough for the sort of server setup you seem to be considering.
>
You've guessed about right. I don't generally do graphics intensive things
such as watching video or gaming, but the machine will need to do a fair
bit of non-graphical stuff. 

Let me see, its normal workload is:
-Apache (internal LAN-only web server) 
-DNS for my LAN
-Postfix central mail-handling for my LAN
-NTP service for the LAN
-LAN-wide backups (rsync)
-nightly local backup (tar to USB)
-Samba (when ever I fire up a 'doze box, increasingly rarely)
-BOINC (seti@HOME + MalariaControl)
-development stuff (C and Java)
-Postgresql (slowly rising load)
-OpenOffice stuff


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

On Sun, 18 May 2008 10:14:31 +0000, Gordon Henderson wrote:

> 
> I suspect I'd look for USB serial/parallel devices these days... Not
> sure if the JAB box will work over one though - maybe I'll have a go
> with mine one day, as it's curently gathering dust...
>
Normally I'd agree, but that won't work with the JABbox, which
sends incoming bits in real time by jiggling the DCD line
rather than sending bytes. Somehow I don't think that would get through a
serial-USB converter though I might be surprised.

If you have a spare Parallax BS1 or BS2 STAMP lurking around the place I
bet that could be persuaded to convert DCD wiggling into bytes. OTOH it
might be easier to simply connect up a blind GPS, read the incoming
GPRMC sentences and extract date&time from them.

> This is one of my older dev. boxes: (About 5 years old now)
>
Many thanks for the CPU details.
 

-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

On Sun, 18 May 2008 11:19:22 +0100, Daniel James wrote:

> In article news:<pan.2008.05.17.22.53.14.13291@see.sig.for.address>, 
> Martin Gregorie wrote:
>> I like the look of the Morex Venus 669 case:
>> http://www.mini-itx.com/store/?c=3#venus
> 
> I have the earlier version of the Venus case ... it's nicely put 
> together but the fans are bloody noisy.
> 
Fair comment, but you haven't heard my OS-9/68000 box! It howls. 
The original, tolerably quiet 80 mm fan in its PSU failed, so I replaced
it with one from my spares box. This runs much faster and sounds more like
a hair dryer. Its definitely louder than my LJ5 in full flow. Fortunately,
all my computers apart from this 'ere ancient Thinkpad live upstairs out
of earshot.

Besides, with one of the lower power mini-ITX boards installed is there
any reason I couldn't disable a fan or two?


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

In article <pan.2008.05.18.16.43.29.770762@see_sig_for_address.invalid>,
Martin Gregorie  <martin@see_sig_for_address.invalid> wrote:
>On Sun, 18 May 2008 11:19:22 +0100, Daniel James wrote:
>
>> In article news:<pan.2008.05.17.22.53.14.13291@see.sig.for.address>, 
>> Martin Gregorie wrote:
>>> I like the look of the Morex Venus 669 case:
>>> http://www.mini-itx.com/store/?c=3#venus
>> 
>> I have the earlier version of the Venus case ... it's nicely put 
>> together but the fans are bloody noisy.
>> 
>Fair comment, but you haven't heard my OS-9/68000 box! It howls. 
>The original, tolerably quiet 80 mm fan in its PSU failed, so I replaced
>it with one from my spares box. This runs much faster and sounds more like
>a hair dryer. Its definitely louder than my LJ5 in full flow. Fortunately,
>all my computers apart from this 'ere ancient Thinkpad live upstairs out
>of earshot.
>
>Besides, with one of the lower power mini-ITX boards installed is there
>any reason I couldn't disable a fan or two?

The 1GHz ones I use for PBXs don't have any fans in them (well - the 1U
rack cases do have a PSU fan)

The only thing to be concerned about is enough ventilation to keep the
disks cool.

I have some EK1000 motherboards working as routers - 2 on-board Ethernet
ports, but this CPU does have a fan, however it's temperature controlled
and I never managed to make it come on during testing.

(Hm. 1 serial, 1 parallel, 2 x sata, 2 x ide ... )

http://linitx.com/viewproduct.php?prodid=11133

Only 64KB cache though! Not sure I'd use it as a router, and all the
other stuff you mentioned in another posting...

Gordon

On Sun, 18 May 2008 17:46:55 +0000, Gordon Henderson wrote:

> The only thing to be concerned about is enough ventilation to keep the
> disks cool.
>
That case I mentioned has a PSU fan, so with luck that will shift enough
air to keep a lite CPU cool. We will see.

> I have some EK1000 motherboards working as routers - 2 on-board
Ethernet
> ports, but this CPU does have a fan, however it's temperature controlled
> and I never managed to make it come on during testing.
> 
> (Hm. 1 serial, 1 parallel, 2 x sata, 2 x ide ... )
> 
> http://linitx.com/viewproduct.php?prodid=11133
> 
> Only 64KB cache though! Not sure I'd use it as a router, and all the
> other stuff you mentioned in another posting...
>
Maybe it has a big brother....

This is all useful input. Thanks. I'm fairly busy at present, but the
contract ends at the end of the coming week, so I'll have time to do some
serious research into mini-ITX boards and get something up and running
from Tuesday week. I'll let you all know how it goes around then.


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

"Nick Kew"  wrote in message 
news:dmp3g5-faq.ln1@grimnir.webthing.com...
> On Sat, 17 May 2008 15:59:35 +0100
> Geoffrey Clements  wrote:
>
>> Martin,
>>
>> I've been thinking about doing this for ages to replace my ageing
>> web-facing server, let us know how you get on.
>>
>> Gordon,
>>
>> thanks for the information, I've been browsing the mini-itx.com
>> web-site and hadn't seen the linitx one. It's very handy to get
>> recommendations from people who have actually used the service.
>>
> I bought from linitx, and was seriously disappointed.
>
> First, I specifically asked about a quiet box (yes, I phoned).
> The case I got was bloody noisy.
>
> Second, I had constant hardware trouble.  The onboard networking
> died fairly quickly, so I had to put a network card in the
> single PCI slot.  Then the PCI slot started playing sillybuggers.
> Then the whole thing started powering down at random and without
> warning, until it failed altogether.
>
> I replaced it with a conventional Sempron-based box based on
> cheapest-available components from DABS (together with the
> disc and memory from the linitx, which are bog-standard
> components and are fine).  It's quieter, and has yet to
> give hardware trouble - apart from being disinclined to work
> with some of my older PCI cards.
>

Cheers Nick,

did you use the returns service at all? I find that a good test of a 
supplier is how efficiently they handle returns.

-- 
Geoff

In article news:<pan.2008.05.18.16.11.13.41188@see_sig_for_address.invalid>, 
Martin Gregorie wrote:
> Are you meaning Celerion and equivalent mainstream laptop CPUs here?

I'm not sure about laptop Celerons, but I think they compare with full-size 
laptop CPUs in roughly the same way that desktop Celerons compare with the 
equivalent full-size desktop CPUs.

When I said that a C3/C7 CPU would give similar performance to a Celeron I 
was thinking of a desktop Celeron, yes.

> If so, how do their power consumption and performance compare with a
> C7?

The C7 barely gets into double figures (of Watts) they're the lowest power 
x86 CPUs around. Some of the Core 2 laptop Celerons are pretty economical 
(and fairly slow) but I don't think they come close to the C7.

> What is there that's a bit nippier than the Celerion, but still fairly
> easy on power?

Once you start to talk about Celerons you're looking at socketed boards with 
intel chipsets, and higher energy consumption (and potentially performance, 
but also price) all round. Such a board will take any of the core mobile 
chips from the most modest single core Celeron to a "+GHz Core 2 Duo -- at a 
price. You'll get more performance, but you *will* use more power and the 
cost will be at least double that of a VIA solution.

The other approach would be to go for an AMD solution. There are mini-ITX 
boards that use low-power Athlon 64 X2 chips, but those CPUs have a TDP of 
45W so could be quite hungry when working hard. There is also at least one 
board that uses the AMD Geode chip, which is quite low power but is also 
quite slow. I think you'd be better off with a C7.

It all depends really what you intend to do with this box...

> You've guessed about right. I don't generally do graphics intensive things
> such as watching video or gaming, but the machine will need to do a fair
> bit of non-graphical stuff. 
> 
> Let me see, its normal workload is:
> -Apache (internal LAN-only web server) 
> -DNS for my LAN
> -Postfix central mail-handling for my LAN
> -NTP service for the LAN
> -LAN-wide backups (rsync)
> -nightly local backup (tar to USB)
> -Samba (when ever I fire up a 'doze box, increasingly rarely)
> -BOINC (seti@HOME + MalariaControl)
> -development stuff (C and Java)
> -Postgresql (slowly rising load)
> -OpenOffice stuff

My mini-ITX system with its 1GHz C3 Nehemiah CPU can do all of those ... I 
run Gentoo linux on it, and it is *really* slow when compiling a big upgrade 
(even with a faster box acting as a distcc server ... that can only help 
when compiling C) so you may find that a C7 system is slower than you'd like 
for "development stuff". I don't know what you mean by "OpenOffice stuff" 
.. OpenOffice apps will run OK on a C7. A C7 will have no trouble with the 
server-ish stuff.

I wouldn't choose to use my box as my main development workstation (too 
slow, 1GB RAM limit), or for gaming or for video editing. It's more than 
adequate for office apps, email, browsing, etc..

Do you really need to support development and office apps on your 24x7 
server? If not, you might consider a dedicated NAS box instead ... there are 
some that run linux (e.g. the qnap boxes), and others that can be made to 
run FreeNAS instead of their own supplied software (notably the Buffalo 
linkstation range). NAS boxes are designed to run 24x7 so they're built 
around very low power components (often using ARM CPUs) but are quite slow 
-- quite fast enough for the server stuff on your list, though.

Cheers,
 Daniel.

In article 
news:<pan.2008.05.18.16.43.29.770762@see_sig_for_address.invalid>, Martin 
Gregorie wrote:
> Besides, with one of the lower power mini-ITX boards installed is there
> any reason I couldn't disable a fan or two?

I'd hesitate to disable the PSU fan ... I've been thinking about removing 
the two case fans (two!) and maybe putting in a low-speed 12cm fan 
somewhere, but the sound from the PSU is sufficient to make me think that 
might be futile effort.

It's not that the noise is loud in absolute terms -- it's certainly 
quieter than my 8-year-old dual Pentium III box with it's worn and dodgy 
fan bearings -- just that it's a case I had hoped would be near-silent, 
but it proved not to be (and that it sits right in front of me on the 
desk).

As the Venus cases are built to take Flex-ATX motherboards there's more 
room inside than in many mini-ITX cases, so there is some scope for 
fiddling with the fan layout.

Cheers,
 Daniel.

On 2008-05-19, Geoffrey Clements  wrote:

> did you use the returns service at all? I find that a good test of a 
> supplier is how efficiently they handle returns.

I returned a wireless keyboard to them, normally I don't bother if the
item is cheap as it's often more hassle than it's worth, but they sent
me a box inside a box with the address and postage already on it, so I
just slapped the old keyboard inside it and dropped it off at the
post office.  The replacement keyboard arrived at the same time as
the packing box, not bad at all.

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

On Mon, 19 May 2008 11:47:14 +0100, Daniel James wrote:

> Do you really need to support development and office apps on your 24x7 
> server?
>
No - its just ended up that way, because my old IBM 866 MHz NetVista is
the most powerful box I own. I usually use it in headless mode from this
even older Thinkpad.

I think a good way to go may be to use a C7 mini-ITX for all the
network-oriented stuff (mail-related, DNS, Apache), backups and the BOINC
thrashing. This would leave the NetVista for development and office
tasks, to be booted on need. Either way the idea is to get a feel for the
capability of these low power boxes.

Thanks for your suggestions.


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

In <4831e0ba$1$30635$834e42db@reader.greatnowhere.com>
Ian Rawlings  wrote:

> On 2008-05-19, Geoffrey Clements
>  wrote:
> 
> > did you use the returns service at all? I find that a good test of
> > a supplier is how efficiently they handle returns.
> 
> I returned a wireless keyboard to them, normally I don't bother if the
> item is cheap as it's often more hassle than it's worth, but they sent
> me a box inside a box with the address and postage already on it, so I
> just slapped the old keyboard inside it and dropped it off at the
> post office.  The replacement keyboard arrived at the same time as
> the packing box, not bad at all.

They should have sent the replacement keyboard inside the box inside a
box instead of sending two packages!

-- 
TH * http://www.realh.co.uk

In article <Te6dnYELDOrua7LVnZ2dnUVZ8sKlnZ2d@plusnet>,
	Andy Burns wrote:
> On 18/05/2008 09:50, Martin Gregorie wrote:

>> I've always heard Rugby referred to as an MSF60 time source. 

> The callsign of the MSF transmitter changed to GBZ when it "moved" from 
> Rugby to Anthorn, that's all.

Nope. It's still MSF for the time signal.

http://www.npl.co.uk/server.php?show=ConWebDoc.998

GBZ is the 19.6kHz transmission to submarines.

-- 
Paul Martin

On 20 May,  
     Paul Martin  wrote:

> 
> GBZ is the 19.6kHz transmission to submarines.
> 
Has it changed? It used to be 19khz. Isn't it now from the VT site at Skelton
Pastures?. Anthorn, IIRC was originally on 22kHz.

-- 
  BD
  Change lycos to yahoo to reply

On 2008-05-19, Tony Houghton  wrote:

> They should have sent the replacement keyboard inside the box inside a
> box instead of sending two packages!

They sent off the replacement keyboard before figuring out what to do
about getting the old one back it seems, which is fine by me!

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

On Tue, 20 May 2008 00:36:45 +0100, me9@privacy.net wrote:

>> GBZ is the 19.6kHz transmission to submarines.
>  
> Has it changed? It used to be 19khz. Isn't it now from the VT site at 
> Skelton Pastures?. Anthorn, IIRC was originally on 22kHz.

Skelton has a big "new" mast and I believe the home of the VLF submarine 
comms when it moved from near Shrewsbury though I thought it was on 
16kHz... Anthorn is the home of MSF and lots of other HF/MF transmitters.

-- 
Cheers
Dave.

On 20 May,  
     "Dave Liquorice"  wrote:

> 
> Skelton has a big "new" mast and I believe the home of the VLF submarine 
> comms when it moved from near Shrewsbury though I thought it was on 
> 16kHz... Anthorn is the home of MSF and lots of other HF/MF transmitters.
> 
Brain fade there, It was 16kHz, and I think its callsign was GBR not GBZ. I
think the site near shrewsbury (Madely?) was a reserve for Rugby. Anthorn was
originally submarine (VLF) comms. I was not aware of anything at HF/MF from
there.

-- 
  BD
  Change lycos to yahoo to reply

On Sun, 18 May 2008 17:30:49 +0100, Martin Gregorie wrote:

> On Sun, 18 May 2008 10:14:31 +0000, Gordon Henderson wrote:
> 
>> 
>> I suspect I'd look for USB serial/parallel devices these days... Not
>> sure if the JAB box will work over one though - maybe I'll have a go
>> with mine one day, as it's curently gathering dust...
>>
> Normally I'd agree, but that won't work with the JABbox, which
> sends incoming bits in real time by jiggling the DCD line
> rather than sending bytes. Somehow I don't think that would get through a
> serial-USB converter though I might be surprised.

It probably would work, but there are two fairly major caveats. Firstly
the jitter will be much worse than when on a PCI/ISA bus. Second very few
(if any) of the serial to USB devices support the TIOCMIWAIT ioctl so it
would have to drop into polling mode, rather sleeping till the change of
state of the RS232 serial lines generates an interrupt which wakes
the program up. This is also not as accurate, and uses far more CPU time.


JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan (at) buzzard.me.uk
St. Andrews, United Kingdom.

On Sat, 17 May 2008 23:53:14 +0100, Martin Gregorie wrote:

[SNIP]

> The trigger for this interest was being given a pair of Fujitsu 20 GB
> > 2.5"
> drives that came out of Pace HDD VCRs. Are these likely to be always on?
> Is there any way I can easily find out?
> 

Check the model number against the manufactures datasheet. Like I said
very few are, and will die after about a year being on 24x7. You can get
around two if you get them sleeping through the night. It is just not
worth the hassle.

>> 
>> For power I use a picoPSU at 60W. The idea here is that with a 6-26V DC
>> input I can build a battery backed PSU rather than a UPS. The concept is
>> that the a 12V lead acid can provide the power when the mains goes away,
>> without conversion back to mains. That said I have not got around to
>> actually building it yet. The only issue is this version does not
>> produce much on the 12V line so all storage has to be laptop based.
>>
> Good idea. I like the idea of having the battery continuously on,
> especially as it should make a decent buffer against mains
> fluctuations and spikes. You'd really need a dual mode charger (one with
> bulk charge and trickle modes) if it is to do much more than keep up with
> the computer's demand. Otherwise it will just wreck the battery, but I
> expect you know this already. Does anybody know if the Maplins dual mode
> chargers can switch back to bulk charge from trickle mode  without being
> switched off and on again? They're good chargers though not terribly
> efficient judging by their normal operating temperature.
> 

The problem with the trickle chargers is that they don't provide enough
juice to power the system and trickle charge. After some more thought and
some Googling at the weekend, I am thinking of one of these

   http://www.powerstream.com/DC-UPS-1212-5A.htm

Would be just the ticket, and designed for the job. What I would like to
do though is add some sort of battery level detection so I can
automatically shutdown the machine before the battery protection kicks in.
The question is whether to use a simple comparator, battery protection
kicks in at 10V, comparator signals on the serial port at say 10.1V by
twiddling a status line. Or do I do a ADC level detection between 13.1 and
10V for battery capacity, something like a TLC548 with an op-amp level
shifter in front and bit bash the serial port. A bit more complicated but
much more pro :-)

I would also want to assert the ring indicator on mains power, so I can
have "Wake on Ring" set in the BIOS to automatically bring it back when
the mains returns.

Power would come from a toroidal mains transformer, but with centre tapped
rectification, avoiding the second diode voltage drop with bridge
rectification, and using some nice expensive Super Barrier Diodes for
extra efficiency. One does not need a regulated output cause the DC-UPS
and the picoPSU take care of that at silly >95% efficiencies anyway.


>> I used a Pack-Box A4 for the case, with the fan removed (my system is
>> fanless for increased reliability).
>>
> I like the look of the Morex Venus 669 case:
> http://www.mini-itx.com/store/?c=3#venus
> because I don't need the rig to be all that small and I like the
> tidiness of an internal PSU.

Well the picoPSU is internal :-) The idea is for the battery backed PSU to
be in a separate wall mounted box, mounted next to the server. It is all
under the stairs, and pretty tidy anyway.

If I was being really pro I would build it all into a 10" rack cabinet

    http://www.cablemonkey.co.uk/acatalog/cabinet.html

> 
>> Finally I have my own design radio clocks (both
>> MSF and DCF77) hooked up to a serial port. The whole lot is mounted on
>> the wall under the stairs.
>>
> I forgot to mention serial ports. Thanks for the reminder: I need two  of
> those too. One for an MSF  clock (your design!) and one for transferring
> data (devices like a Garmin GPS II+ and an EW flight recorder).
> 

The Via board that I use has one on the ATX IO plate and a header on the
board for another.


JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan (at) buzzard.me.uk
St. Andrews, United Kingdom.

On Tue, 20 May 2008 17:07:50 +0100, me9@privacy.net wrote:

> Brain fade there, It was 16kHz, and I think its callsign was GBR not 
> GBZ.

http://www.subbrit.org.uk/sb-sites/sites/r/rugby_radio/index.shtml

Has Rugby as GBR.

> I think the site near shrewsbury (Madely?) was a reserve for Rugby. 

Not Madly. Criggion, it was fairly recently demolished, one half the 
aerial support system was a hill, and yes it standby for Rugby. (See above 
link)

http://www.subbrit.org.uk/sb-sites/sites/c/criggion_radio/

> Anthorn was originally submarine (VLF) comms. I was not aware of 
> anything at HF/MF from there.

Hum I'm going from the gert big aerial farm they have at Anthorn. I *may 
well* be getting confused with another site but I seem to have a memeory 
from the one time I've driven past of of cicular bird cages and big log 
periodics on masts. It appears I am:

http://www.visitcumbria.com/car/anthorn.htm

Not sure I trust wikipedia for this but it lists GBR as Rugby om 15.8kHz. 
GBZ as Anthorm on 19.6Khz. GBZ is also listed for Skelton on 22.1kHz.

-- 
Cheers
Dave.

On Tue, 20 May 2008 21:07:24 +0100
Jonathan Buzzard  wrote:

> The problem with the trickle chargers is that they don't provide enough
> juice to power the system and trickle charge. After some more thought and
> some Googling at the weekend, I am thinking of one of these

Use a simple basic PSU to provide raw but smoothed power and also
trickle charge the battery via a simple voltage regulator/current
limiter.

Raw power is fed via a diode to a SMPSU to give nice regulated output.

Voltage drop across diode is monitored and fed to the gate of a
P-channel power MOSFET via suitable amplifier/switch. MOSFET is
connected between battery and SMPSU. When mains goes off voltage across
diode drops and FET switches on.

When mains is restored drop across diode reappears and FET is switched
off. Ideally raw power voltage should only be a couple of volts above
fully charged battery voltage.

-- 
Will J G

On Tue, 20 May 2008 19:42:32 +0100, Jonathan Buzzard wrote:

> 
> It probably would work, but there are two fairly major caveats. Firstly
> the jitter will be much worse than when on a PCI/ISA bus. Second very few
> (if any) of the serial to USB devices support the TIOCMIWAIT ioctl so it
> would have to drop into polling mode, rather sleeping till the change of
> state of the RS232 serial lines generates an interrupt which wakes
> the program up. This is also not as accurate, and uses far more CPU time.
>
I've always had to use polling anyway - first with an El Cheapo Lava
serial board and then on the NetVista with whatever it uses for the
on-mobo serial ports.


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

On Tue, 20 May 2008 21:07:24 +0100, Jonathan Buzzard wrote:

> I would also want to assert the ring indicator on mains power, so I can
> have "Wake on Ring" set in the BIOS to automatically bring it back when
> the mains returns.

You may want to have a delay between the mains coming back and stuff 
powering back up to give your battery a chance to recharge a bit. So if 
the power goes again there is guaranteed to be enough juice in the battery 
for an orderly shutdown.

-- 
Cheers
Dave.

On Tue, 20 May 2008 21:07:24 +0100, Jonathan Buzzard wrote:

 
> The problem with the trickle chargers is that they don't provide enough
> juice to power the system and trickle charge. After some more thought and
> some Googling at the weekend, I am thinking of one of these
> 
>    http://www.powerstream.com/DC-UPS-1212-5A.htm
> 
Cute. However, I still think a standard dual mode charger may do the job,
though on second thoughts the Maplins jobbie wouldn't - only 1.5 amps at
14v.

And, as you say, a voltage low detector is still needed.

> Well the picoPSU is internal :-) The idea is for the battery backed PSU
> to be in a separate wall mounted box, mounted next to the server. It is
> all under the stairs, and pretty tidy anyway.
>
Yes - brain fade on my part - IIRC its the cute little thing that plugs
into the mini-ITX power socket with 12v on its inlets.

> The Via board that I use has one on the ATX IO plate and a header on the
> board for another.
>
Another good point for Via boards!


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

On Tue, 20 May 2008 22:28:47 +0100, Folderol wrote:

> On Tue, 20 May 2008 21:07:24 +0100
> Jonathan Buzzard  wrote:
> 
>> The problem with the trickle chargers is that they don't provide enough
>> juice to power the system and trickle charge. After some more thought and
>> some Googling at the weekend, I am thinking of one of these
> 
> Use a simple basic PSU to provide raw but smoothed power and also
> trickle charge the battery via a simple voltage regulator/current
> limiter.

Which would knacker the lifespan of my gel lead acid battery.

> Raw power is fed via a diode to a SMPSU to give nice regulated output.

Except the picoPSU 60W I have is a wide range 6-26V input. It does not need
a regulated input. Smoothed yes, regulated no. It is a switched mode DC-DC
converter itself. I have already tested that with a haxored setup to make
sure it worked.

> Voltage drop across diode is monitored and fed to the gate of a
> P-channel power MOSFET via suitable amplifier/switch. MOSFET is
> connected between battery and SMPSU. When mains goes off voltage across
> diode drops and FET switches on.

Hum, why not power it from the unregulated output of the transformer? That
way I avoid the voltage drop across the diode with associated loss.
Remember a standard diode with a 1.1V drop supplying 3A current at 12V of
power will waste 3.3W That is 10% of my total power budget!!!


> When mains is restored drop across diode reappears and FET is switched
> off. Ideally raw power voltage should only be a couple of volts above
> fully charged battery voltage.
>

But why bother when I can buy a professional solution, with proper three
stage microprocessor charging of the battery (so longer battery life), and
battery deep discharging protection for a reasonable price (32GBP at
current exchange rates)?


JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan (at) buzzard.me.uk
St. Andrews, United Kingdom.

On Tue, 20 May 2008 23:02:10 +0100, Martin Gregorie wrote:

> On Tue, 20 May 2008 19:42:32 +0100, Jonathan Buzzard wrote:
> 
>> 
>> It probably would work, but there are two fairly major caveats. Firstly
>> the jitter will be much worse than when on a PCI/ISA bus. Second very few
>> (if any) of the serial to USB devices support the TIOCMIWAIT ioctl so it
>> would have to drop into polling mode, rather sleeping till the change of
>> state of the RS232 serial lines generates an interrupt which wakes
>> the program up. This is also not as accurate, and uses far more CPU time.
>>
> I've always had to use polling anyway - first with an El Cheapo Lava
> serial board and then on the NetVista with whatever it uses for the
> on-mobo serial ports.
> 
>

Nasty, always use the TIOCMIWAIT myself on several Toshiba laptops and the
VIA board I now use

jab@small:~$ more /proc/interrupts
           CPU0
  0:  132545550    IO-APIC-edge  timer
  4:    2809226    IO-APIC-edge  serial
  8:          1    IO-APIC-edge  rtc
  9:          0   IO-APIC-level  acpi
 14:    1288847    IO-APIC-edge  ide0
 15:    1288841    IO-APIC-edge  ide1
169:          2   IO-APIC-level  ohci1394
177:          0   IO-APIC-level  uhci_hcd:usb1, uhci_hcd:usb2, uhci_hcd:usb3, ehci_hcd:usb4
185:     122918   IO-APIC-level  eth1
201:   20342870   IO-APIC-level  unicorn_pci
NMI:          0
LOC:  132545493
ERR:          0
MIS:          0

Only been up 6 days due to kernel upgrade on Debian. Usually that serial
number is huge.


jab@small:~$ ntpq -p
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 LOCAL(1)        .LOCL.          10 l   62   64  377    0.000    0.000   0.001
*SHM(0)          .MSF.            0 l   75   64  332    0.000   -0.718   1.009
 SHM(1)          .DCF7.           0 l 107m   64    0    0.000   -6.283   0.001

Hum, looks like the DCF77 receiver has fallen off again :-( Need to work
out something better than BlueTak me thinks.


JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan (at) buzzard.me.uk
St. Andrews, United Kingdom.

On Tue, 20 May 2008 23:51:01 퍝
Jonathan Buzzard  wrote:

> But why bother when I can buy a professional solution, with proper three
> stage microprocessor charging of the battery (so longer battery life), and
> battery deep discharging protection for a reasonable price (32GBP at
> current exchange rates)?

From your original post I got the impression you liked getting your
hands dirty and experimenting.

I was obviously wrong <shrug>

-- 
Will J G

On 2008-05-20, Folderol  wrote:

> From your original post I got the impression you liked getting your
> hands dirty and experimenting.
>
> I was obviously wrong <shrug>

There's a difference between liking it and insisting on it ;-)

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

On Tue, 20 May 2008 22:27:32 +0100, Dave Liquorice wrote:

> On Tue, 20 May 2008 21:07:24 +0100, Jonathan Buzzard wrote:
> 
>> I would also want to assert the ring indicator on mains power, so I can
>> have "Wake on Ring" set in the BIOS to automatically bring it back when
>> the mains returns.
> 
> You may want to have a delay between the mains coming back and stuff 
> powering back up to give your battery a chance to recharge a bit. So if 
> the power goes again there is guaranteed to be enough juice in the battery 
> for an orderly shutdown.
>

Good point. A simple level detection on the battery voltage, that triggers
at a higher level than the shutdown, when mains is present should do the
trick I think.


JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan (at) buzzard.me.uk
St. Andrews, United Kingdom.

On Wed, 21 May 2008 07:05:57 +0100, Ian Rawlings wrote:

> On 2008-05-20, Folderol  wrote:
> 
>> From your original post I got the impression you liked getting your
>> hands dirty and experimenting.
>>
>> I was obviously wrong <shrug>
> 
> There's a difference between liking it and insisting on it ;-)
>

Indeed. Also good design is about making innovative use of preexisting
building blocks. I will be getting my hands dirty, I am doing my own mains
to DC, and adding some custom components to the battery side to signal
stuff to the computer to make it all work in a pro fashion.


JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan (at) buzzard.me.uk
St. Andrews, United Kingdom.

On Wed, 21 May 2008 21:42:58 +0100, Jonathan Buzzard wrote:

>> You may want to have a delay between the mains coming back and stuff 
>> powering back up to give your battery a chance to recharge a bit. 
> 
> Good point. A simple level detection on the battery voltage, that 
> triggers at a higher level than the shutdown, when mains is present 
> should do the trick I think.

If you can distinguish between battery voltage and voltage from the 
charger...  Might be easier just to have a timed delay worked out from how 
long it should take the charger to put back about 15% capacity.

-- 
Cheers
Dave.

On Sun, 18 May 2008 17:43:30 +0100, Martin Gregorie wrote:

> Besides, with one of the lower power mini-ITX boards installed is there
> any reason I couldn't disable a fan or two?

Part of lm-sensors [round here at least] is something called 'pwmconfig' 
which one runs to configure the 'fancontrol' service. With any luck, the 
fan(s) in your new box will be supported by lm-sensors, and the fans will 
be varied in speed automatically by the system.

-- 
 <http://ale.cx/> (AIM:troffasky) (UnSoEsNpEaTm@ale.cx)
 22:04:06 up  1:57,  1 user,  load average: 0.03, 0.04, 0.16
 Convergence, n: The act of using separate DSL circuits for voice and data

On Sat, 17 May 2008 17:48:51 +0100, Jonathan Buzzard wrote:

> 
> For hard drives a use a couple of "Always-On" rated Hitachi 2.5" 7200RPM
> IDE drives in software RAID-1. I cannot stress how important it is to get
> the "Always-On" versions of the drives.
>
Many thanks for that advice. The mini-ITX build and feedback have been
delayed due to a contract extension, so nothing to report yet.

However I did look up the spec of the Fujitsu 2.5" drives, a MHT2020AT and
a MHS2020AT. Now I fully understand Jonathon's warnings about only using
an always-on drive - these drives are only warranted for being
powered up for 250 hours per month or less and should have less than a 20%
duty cycle during the power on time. IOW, their expected life is 9000
hours over 3 years of 300,000 start/stop cycles.

Looks like a standard 3.5" or a so-called 2.5" Enterprise drive is the way
to go for my requirements.  


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

On Tue, 03 Jun 2008 22:02:11 퍝, Martin Gregorie wrote:

> However I did look up the spec of the Fujitsu 2.5" drives, a MHT2020AT> and a MHS2020AT. Now I fully understand Jonathon's warnings about only> using an always-on drive - these drives are only warranted for being
> powered up for 250 hours per month or less and should have less than a> 20% duty cycle during the power on time. IOW, their expected life is 
> 9000 hours over 3 years of 300,000 start/stop cycles.

Flippin heck! I've just bought a 3.5" 160G Seagate ST3160215A for <£40with a MTBF of 700,000 hours (a tad under 80 years) and annual failure 
rate of 0.34%.

-- 
Cheers
Dave.

In article <pan.2008.06.03.21.02.10.616837@see_sig_for_address.invalid>,
	Martin Gregorie wrote:
> However I did look up the spec of the Fujitsu 2.5" drives, a MHT2020AT and
> a MHS2020AT. Now I fully understand Jonathon's warnings about only using
> an always-on drive - these drives are only warranted for being
> powered up for 250 hours per month or less and should have less than a 20%
> duty cycle during the power on time. IOW, their expected life is 9000
> hours over 3 years of 300,000 start/stop cycles.

Mine's (Hitachi Travelstar) slightly better:

 333 hours per month,
 20% duty cycle,
 first of 20,000 hours or 5 years,
 600,000 head load/unloads

-- 
Paul Martin

On Wed, 04 Jun 2008 01:20:00 +0100, Paul Martin wrote:

> 
> Mine's (Hitachi Travelstar) slightly better:
> 
>  333 hours per month,
>  20% duty cycle,
>  first of 20,000 hours or 5 years,
>  600,000 head load/unloads
>
Unsurprisingly, the drives I was given are not current models. I did
glance at the specs for a current Fujitsu while trying to find specs for
an "always spinning" 2.5" drive. The two figures that stuck in my mind
were 600,000 start/stop cycles and 1.8 watts. Looks like the Hitachi's
figures are fairly typical.

Is it a safe guess that the only "always spinning" 2.5" drives are the
so-called "Enterprise 2.5" units? If so, Enterprise 2.5" or 3.5" the 
ones to use for anything other than fairly light use.

That said, my old Thinkpad has run more like 10 hours a day, 5 days a week
for the last four years, without any disk problems. As it only seems to
spin the drive down when its on battery, its drive has certainly now
exceeded the Fujitsu figures by a handsome margin: around 10,500 hours at
100% duty cycle in the last four years alone - and I've used it for 5
years before that. 

FWIW, as the Fujitsus came out of a HDD VCR, thought about the likely duty
cycle. In that task the VCR could turn the drives off completely unless
its recording a show or being watched, when the drives would be spun up
throughout the task. I guessed that a VCR might be recording or
being watched about 50 hours per month which, coincidentally, is close to
250 hours/month at 20% duty cycle.


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

In <pan.2008.06.04.18.39.45.27877@see_sig_for_address.invalid>
Martin Gregorie <martin@see_sig_for_address.invalid> wrote:

> FWIW, as the Fujitsus came out of a HDD VCR, thought about the likely
> duty cycle. In that task the VCR could turn the drives off completely
> unless its recording a show or being watched, when the drives would
> be spun up throughout the task. I guessed that a VCR might be
> recording or being watched about 50 hours per month which,
> coincidentally, is close to 250 hours/month at 20% duty cycle.

By "duty cycle" are we talking about the drives being spun down when off
duty or just idly spinning without being accessed?

-- 
TH * http://www.realh.co.uk

On Wed, 04 Jun 2008 21:28:12 +0100, Tony Houghton wrote:

> By "duty cycle" are we talking about the drives being spun down when off
> duty or just idly spinning without being accessed?
>
It actually says:

	Operating time: 20% or less of power-on time

so I assume that means that for 20% of the time its powered up the disk is
spinning and that its spun down for 80% of the time. They don't explain
what they mean by 'operating time'. 

BTW I picked up the design life of 3 years in another part of the site.
In the manual it says the MTBF is 300,000 hours provided that:
	- power-on time is less than 250 hrs/mth or 3000 hrs/yr
 	- operating time 20% or less of power on time
	- environment 5-55C, 8-90% humidity 
	  but humidity bulb thermometer <=29C

and quotes 5 years/ 20,000 hours between services if its run below 48C.


-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

On Wed, 04 Jun 2008 19:39:45 +0100, Martin Gregorie wrote:

> On Wed, 04 Jun 2008 01:20:00 +0100, Paul Martin wrote:
> 
>> 
>> Mine's (Hitachi Travelstar) slightly better:
>> 
>>  333 hours per month,
>>  20% duty cycle,
>>  first of 20,000 hours or 5 years,
>>  600,000 head load/unloads
>>
> Unsurprisingly, the drives I was given are not current models. I did
> glance at the specs for a current Fujitsu while trying to find specs for
> an "always spinning" 2.5" drive. The two figures that stuck in my mind
> were 600,000 start/stop cycles and 1.8 watts. Looks like the Hitachi's
> figures are fairly typical.

Note that as far as I am aware only Hitachi sell 2.5" drives with a
7200rpm spindle speed, that are "always on" rated.

> Is it a safe guess that the only "always spinning" 2.5" drives are the
> so-called "Enterprise 2.5" units? If so, Enterprise 2.5" or 3.5" the 
> ones to use for anything other than fairly light use.
> 
> That said, my old Thinkpad has run more like 10 hours a day, 5 days a week
> for the last four years, without any disk problems. As it only seems to
> spin the drive down when its on battery, its drive has certainly now
> exceeded the Fujitsu figures by a handsome margin: around 10,500 hours at
> 100% duty cycle in the last four years alone - and I've used it for 5
> years before that. 

I have been running low powered home servers i.e. less than 30W, for over
eight years now. First using laptops and for the last 18 months as a
mini-ITX solution. I had three drive failures in the first four and a half
years. At which point I got sick, decided to pony up the extra cash for a
60GB "always on" 7200rpm Hitachi unit. While it was retired from the
server after three years (capacity issues and a desire to go to RAID1) it
is still going strong in a 2.5" USB drive enclosure.

I freely admit that my sample size is to small, but the life spans I was
getting matched closely with the data sheets. For me the small extra cost
is worth the decrease in hassle.

I would also add that my home server does everything, the broadband, WiFi
access point, mail server, file server etc. When it goes south it is a
major pain, so anything that reduces this likely hood without spending too
much is worth the money.
 
> FWIW, as the Fujitsus came out of a HDD VCR, thought about the likely duty
> cycle. In that task the VCR could turn the drives off completely unless
> its recording a show or being watched, when the drives would be spun up
> throughout the task. I guessed that a VCR might be recording or
> being watched about 50 hours per month which, coincidentally, is close to
> 250 hours/month at 20% duty cycle.
> 

I bet it would have a fairly aggressive spin down schedule. That does make
them last longer.


JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan (at) buzzard.me.uk
St. Andrews, United Kingdom.

On Thu, 05 Jun 2008 00:49:53 +0100, Jonathan Buzzard wrote:

> Note that as far as I am aware only Hitachi sell 2.5" drives with a
> 7200rpm spindle speed, that are "always on" rated.
>
Fujitsu sell 7200 rpm 2.5's under the "Enterprise 2.5" label, which is why
I phrased the question the way I did.
 
> I freely admit that my sample size is to small, but the life spans I was
> getting matched closely with the data sheets. For me the small extra
> cost is worth the decrease in hassle.
> 
There are a lot of things that can provoke this. I had a spate of using
Maxstor drives, because some of their old ones (we're talking sub 1GB
here) were good. Then I realised I was having a drive die every couple of
years or so and switched to WD. No failures from them yet - though as the
last Maxstor died in late 2005 its a bit early to tell yet...

> I bet it would have a fairly aggressive spin down schedule. That does
> make them last longer.
>
I can see no reason for the drives in a VCR to be even powered on unless
the machine is actually recording a program or playing one back. I got
to 50 hours/month be using that assumption together with a guess that the
machine would record one 30 minute program a day and that the previous
day's program would be watched and then adding a 20 hour margin. 

I don't have TV, so have no idea how much a VCR gets in reality. Its
recording + replay time could easily be quite a bit less than my
guestimate.
 

-- 
martin@   | Martin Gregorie
gregorie. | 
org       | Zappa fan & glider pilot

In message <pan.2008.06.06.23.34.00.385466@see_sig_for_address.invalid>, 
Martin Gregorie <martin@see_sig_for_address.invalid> writes
>
>I can see no reason for the drives in a VCR to be even powered on unless
>the machine is actually recording a program or playing one back.
>
PVRs (at least the ones I am familiar with) have the drives on and 
working continuously when the box is powered up because they are 
buffering live TV to allow pause and rewind of the channel(s) being 
watched.  Our PVR is probably powered up 450 hours/month.
-- 
Martin Liddle, Tynemouth Computer Services, 3 Kentmere Way,
Staveley, Chesterfield, Derbyshire, S43 3TW.
Web site: <http://www.tynecomp.co.uk>.

On 2008-06-07, Martin Liddle  wrote:

> PVRs (at least the ones I am familiar with) have the drives on and 
> working continuously when the box is powered up because they are 
> buffering live TV to allow pause and rewind of the channel(s) being 
> watched.  Our PVR is probably powered up 450 hours/month.

No point buffering live TV that no-one's watching, otherwise you're
just wearing out the hard disc extremely prematurely by not only
having it powered on, but continuously writing.  My last PVR (a
home-made MythTV box) only buffered TV when I was watching it, so it
might be worth checking that yours actually is buffering live TV when
you're not watching it, chances are it's not, and if it is, you may
well be forgetting to do something to take it out of live TV mode.

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

In article news:<WmZDaWnTokSIFwKa@tynecomp.invalid>, Martin Liddle wrote:
> PVRs (at least the ones I am familiar with) have the drives on and 
> working continuously when the box is powered up because they are 
> buffering live TV to allow pause and rewind of the channel(s) being 
> watched.

What make?

Mine (Toshiba) doesn't buffer live TV unless and until you press the "Pause" button while watching live TV. Then it buffers until you stop it.

I'm not sure whether the HDD is powered when not in use, though. I suppose the PVR could contain a relay to power the drive down when not in use but I'd be surprised if it does. These drives do have very low-power sleep modes, and I'd expect the drives to be in that state when the device is on but the disk is not operating.

This is quite interesting reading - I see nothing here to suggest that the disks shouldn't be on 24/7 if the application required it:
http://www.seagate.com/staticfiles/support/disc/manuals/ce/DB35%20Series/DB35.2%20Series/100390003b.pdf

The trouble with technology is that the cleverer it gets the cleverer you expect it to be ... I am forever pressing "fast forward" while watching live TV to try to skip over the adverts to something that hasn't even been broadcast yet!

Cheers,
 Daniel.

In message , Daniel James 
 writes
>In article news:<WmZDaWnTokSIFwKa@tynecomp.invalid>, Martin Liddle wrote:
>> PVRs (at least the ones I am familiar with) have the drives on and
>> working continuously when the box is powered up because they are
>> buffering live TV to allow pause and rewind of the channel(s) being
>> watched.
>
>What make?
>
Humax PVR 9200
-- 
Martin Liddle, Tynemouth Computer Services, 3 Kentmere Way,
Staveley, Chesterfield, Derbyshire, S43 3TW.
Web site: <http://www.tynecomp.co.uk>.

In message <484a5a5b$0$30637$834e42db@reader.greatnowhere.com>, Ian 
Rawlings  writes
>
>No point buffering live TV that no-one's watching, otherwise you're
>just wearing out the hard disc extremely prematurely by not only
>having it powered on, but continuously writing.  My last PVR (a
>home-made MythTV box) only buffered TV when I was watching it, so it
>might be worth checking that yours actually is buffering live TV when
>you're not watching it, chances are it's not, and if it is, you may
>well be forgetting to do something to take it out of live TV mode.
>
I am quite sure that it is continuously buffering (and we find this a 
useful facility)and there is no option to turn it off.
-- 
Martin Liddle, Tynemouth Computer Services, 3 Kentmere Way,
Staveley, Chesterfield, Derbyshire, S43 3TW.
Web site: <http://www.tynecomp.co.uk>.

On 2008-06-07, Martin Liddle  wrote:

> I am quite sure that it is continuously buffering (and we find this a 
> useful facility)and there is no option to turn it off.

When you are watching TV of course it's continuously buffering, as
does mythtv, but when you are *not* watching TV?  When the TV is
turned off and you're not at home, it's still buffering?

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

In message , Ian Rawlings 
 writes
>On 2008-06-07, Martin Liddle  wrote:
>
>> I am quite sure that it is continuously buffering (and we find this a
>> useful facility)and there is no option to turn it off.
>
>When the TV is turned off and you're not at home, it's still buffering?
>
If the PVR is on then yes it is buffering regardless of the state of the 
TV.  The PVR is turned off when no one is home (and overnight) but as I 
work from home it is on most days.
-- 
Martin Liddle, Tynemouth Computer Services, 3 Kentmere Way,
Staveley, Chesterfield, Derbyshire, S43 3TW.
Web site: <http://www.tynecomp.co.uk>.

Ian Rawlings  writes:

> On 2008-06-07, Martin Liddle  wrote:
>
>> I am quite sure that it is continuously buffering (and we find this a 
>> useful facility)and there is no option to turn it off.
>
> When you are watching TV of course it's continuously buffering, as
> does mythtv, but when you are *not* watching TV?  When the TV is
> turned off and you're not at home, it's still buffering?

Mine is (a Humax).  I find this feature more useful than if it did
only as you describe.  Reviewing something I have (or someone else
has) been watching is used less that switching on and flicking back to
that start of something I'd not set to record.  No doubt this just
reflects one particular use pattern.  I'll let you know whether I
think it is worth the reduced disc/head life when the thing dies :-)

-- 
Ben.

On 2008-06-07, Martin Liddle  wrote:

> If the PVR is on then yes it is buffering regardless of the state of the 
> TV.  The PVR is turned off when no one is home (and overnight) but as I 
> work from home it is on most days.

Hmm, on my PVR the PVR was on at all times so it could record
scheduled programmes, but it had a Live TV mode that I could turn to
to watch live TV, once in that mode it would start buffering but when
not in live TV it would no longer buffer but would instead sit there
updating TV schedules, recording programmes at set times and when it
wasn't doing those two, it would power off the hard disc.  That way
the hard disc lasts much longer and the unit draws less power.  Either
yours has a broken design or you're leaving it in live TV mode all the
time from the sound of it.

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

On 2008-06-07, Ben Bacarisse  wrote:

> Mine is (a Humax).  I find this feature more useful than if it did
> only as you describe.  Reviewing something I have (or someone else
> has) been watching is used less that switching on and flicking back to
> that start of something I'd not set to record.  No doubt this just
> reflects one particular use pattern.  I'll let you know whether I
> think it is worth the reduced disc/head life when the thing dies :-)

Being able to flick back only works on the channel that it happens to
have been left on surely?  Or does it have multiple tuners that allows
it to buffer a range of channels at once?

MythTV could do what you describe if left in live TV mode, but given
the hard disc wear and extra power draw, and the few times I'd find it
useful I personally didn't bother.  What I did instead was fit a 500
gig disc and set it to record just about anything that there was a
small chance I might watch, with programmes set to auto-delete when
the disc was full, certain programmes would be set to never
auto-delete so the majority of programmes would just expire unwatched
and the ones I really wanted to watch would stay on the disc until I
watched them and manually deleted them.  It used to record about 6-8
hours of TV per day, of that I'd watch maybe 1 hour, sometimes when
working I'd leave it to play a load of programmes from up to a few
months back in the background.  I used to record all the old movies
from the 60s and other dross from daytime TV and play it back in a
window on my desktop  while working.  

These days I don't even have TV, no aerial on the roof, and download
the Formula One races and Have I Got News For You, the only things I
ever really watched properly anyway ;-)

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

Ian Rawlings  writes:

> On 2008-06-07, Ben Bacarisse  wrote:
>
>> Mine is (a Humax).  I find this feature more useful than if it did
>> only as you describe.  Reviewing something I have (or someone else
>> has) been watching is used less that switching on and flicking back to
>> that start of something I'd not set to record.  No doubt this just
>> reflects one particular use pattern.  I'll let you know whether I
>> think it is worth the reduced disc/head life when the thing dies :-)
>
> Being able to flick back only works on the channel that it happens to
> have been left on surely?

Yes.

>  Or does it have multiple tuners that allows
> it to buffer a range of channels at once?

It has multiple tuners (2) but only one is used to do this buffering.
The hardware can record two streams at once (if they are on different
channels) and can decode and play a third at the same time, but
presumably they though that having two channels buffered would present
users with some confusion.

-- 
Ben.

On Sat, 7 Jun 2008 10:40:25 +0100
Ian Rawlings  wrote:

> No point buffering live TV that no-one's watching, otherwise you're
> just wearing out the hard disc extremely prematurely by not only
> having it powered on, but continuously writing.  My last PVR (a
> home-made MythTV box) only buffered TV when I was watching it, so it
> might be worth checking that yours actually is buffering live TV when
> you're not watching it, chances are it's not, and if it is, you may
> well be forgetting to do something to take it out of live TV mode.

The trouble with that is the default state of a typical STB is to be
showing a channel and it doesn't really have any way of knowing whether
the TV is actually switched on and being watched.

With RAM so cheap now a good strategy would be to load the box up with
RAM and use that for buffering.

-- 
TH * http://www.realh.co.uk

On Sat, 7 Jun 2008 13:39:18 +0100
Ian Rawlings  wrote:

> These days I don't even have TV, no aerial on the roof, and download
> the Formula One races and Have I Got News For You, the only things I
> ever really watched properly anyway ;-)

Oh, you can download F1? That's good news. There's no way it's going to
go back to BBC when ITV's contract expires. Sky will snap it up "at the
last moment".

Oops, getting really OT.

-- 
TH * http://www.realh.co.uk

On Sat, 7 Jun 2008 10:40:25 +0100, Ian Rawlings wrote:

> My last PVR (a home-made MythTV box) only buffered TV when I was 
> watching it, 

How does it know if you are watching it?

My Philips DVD/HDD recoreder buffered the last 3 hours of what was on it's 
output regardless of any one watching it. Very handy and easy to use. The 
Sony that replaced the Phillips (as it failed twice under warranty) may 
also do the same but the user interface is a load of cack compared to the 
Philips so I haven't worked out how to pause/rewind live TV. It's not just 
the pause or rewind buttons...

-- 
Cheers
Dave.

On 2008-06-07, Tony Houghton  wrote:

> Oh, you can download F1? That's good news. There's no way it's going to
> go back to BBC when ITV's contract expires. Sky will snap it up "at the
> last moment".

As far as I'm aware, the BBC has already got it sewn up, there's never
been any mention of sky as ever having been interested, F1 is the poor
relation to football, the BBC lost football so went for F1 instead,
and probably world curling championships, plus european pro poosticks.

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

In message , Ian Rawlings 
 writes
>Either yours has a broken design or you're leaving it in live TV mode 
>all the time from the sound of it.
>
There is no way to turn off buffering mode.  The general design 
philosophy of the Humax 9200 is to have a minimal set of options and for 
my technophobic partner that works very well.
-- 
Martin Liddle, Tynemouth Computer Services, 3 Kentmere Way,
Staveley, Chesterfield, Derbyshire, S43 3TW.
Web site: <http://www.tynecomp.co.uk>.

On 2008-06-07, Dave Liquorice  wrote:

> On Sat, 7 Jun 2008 10:40:25 +0100, Ian Rawlings wrote:
>
>> My last PVR (a home-made MythTV box) only buffered TV when I was 
>> watching it, 
>
> How does it know if you are watching it?

'cos I'd tell it!

Basically the mythtv setup used the server upstairs that did the
recording, and the clients around the house did the displaying and
user interface.  If I wasn't in live TV mode, or wasn't running a
client, then it didn't buffer live TV.  My lounge "TV" is just a big
LCD screen, so if I wanted to watch telly I'd fire up the myth
frontend from a Gnome menu and go to live TV mode, and that would then
start the buffering.

> My Philips DVD/HDD recoreder buffered the last 3 hours of what was on it's 
> output regardless of any one watching it. Very handy and easy to use.

Well yeah but sort of like throwing a cloth over the TV when you're
not watching it ;-)

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

On 2008-06-07, Tony Houghton  wrote:

> The trouble with that is the default state of a typical STB is to be
> showing a channel and it doesn't really have any way of knowing whether
> the TV is actually switched on and being watched.

No but it would be easy to have the STB, when in standby mode, wake up
on a timer to record a given programme so the user can just put it
into standby mode (now the hysteria about standby modes founded on
inaccurate figures has passed) when it's not in use.  No way to
compensate for daft users just leaving it running of course ;-)

-- 
Blast off and strike the evil Bydo empire!
http://youtube.com/user/tarcus69
http://www.flickr.com/photos/tarcus/sets/

In 
Ian Rawlings  wrote:

> On 2008-06-07, Tony Houghton  wrote:
> 
> > Oh, you can download F1? That's good news. There's no way it's
> > going to go back to BBC when ITV's contract expires. Sky will snap
> > it up "at the last moment".
> 
> As far as I'm aware, the BBC has already got it sewn up, there's never
> been any mention of sky as ever having been interested, F1 is the poor
> relation to football, the BBC lost football so went for F1 instead,
> and probably world curling championships, plus european pro poosticks.

At least they're not desperate enough to go for A1GP ;-).

-- 
TH * http://www.realh.co.uk

In message , Jonathan Buzzard 
 writes
>
>I have been running low powered home servers i.e. less than 30W, for over
>eight years now. First using laptops and for the last 18 months as a
>mini-ITX solution. I had three drive failures in the first four and a half
>years. At which point I got sick, decided to pony up the extra cash for a
>60GB "always on" 7200rpm Hitachi unit. While it was retired from the
>server after three years (capacity issues and a desire to go to RAID1) it
>is still going strong in a 2.5" USB drive enclosure.
>
How viable are solid state IDE drives for this sort of application?  I 
have just had a 2.5" conventional IDE drive fail and I was wondering 
whether it made sense to replace it with a solid state IDE drive.  4GB 
is probably big enough for my application.  Any recommendations on 
brands or suppliers?
-- 
Martin Liddle, Tynemouth Computer Services, 3 Kentmere Way,
Staveley, Chesterfield, Derbyshire, S43 3TW.
Web site: <http://www.tynecomp.co.uk>.

In article <2QZrX2MUwoTIFwFs@tynecomp.invalid>,
Martin Liddle   wrote:
>In message , Jonathan Buzzard 
> writes
>>
>>I have been running low powered home servers i.e. less than 30W, for over
>>eight years now. First using laptops and for the last 18 months as a
>>mini-ITX solution. I had three drive failures in the first four and a half
>>years. At which point I got sick, decided to pony up the extra cash for a
>>60GB "always on" 7200rpm Hitachi unit. While it was retired from the
>>server after three years (capacity issues and a desire to go to RAID1) it
>>is still going strong in a 2.5" USB drive enclosure.
>>
>How viable are solid state IDE drives for this sort of application?  I 
>have just had a 2.5" conventional IDE drive fail and I was wondering 
>whether it made sense to replace it with a solid state IDE drive.  4GB 
>is probably big enough for my application.  Any recommendations on 
>brands or suppliers?

If you want a "live" filesystem, then the bigger the better for their
"wear-levelling" features - so that basically means one large / partition.
You also might want to use ext2, or turn the options on to increase the
ext3 commit time (fstab options: defaults,noatime,commit=43200)

I build my mini ITX system booting from flash IDE drive (32 or 64MB), but
they all run in RAM after that - so the flash disk sits unused 99% of the
time, however I'm looking at 4 or 8GB flash IDE drives for my desktop and
development boxes (with /home NFS mounted off the main server elsewhere)

I get my ones from:

  http://linitx.com/viewcategory.php?catid=129&pp=100,129

They're not cheap though - the best part of 100 quid for the 8GB
modules. (er, + VAT so more than 100!)

Gordon

"Gordon Henderson" <gordon+usenet@drogon.net> wrote in message 
news:g2m8mv$2f9m$1@energise.enta.net...
> In article <2QZrX2MUwoTIFwFs@tynecomp.invalid>,

[snip]

> I build my mini ITX system booting from flash IDE drive (32 or 64MB), but
> they all run in RAM after that - so the flash disk sits unused 99% of the
> time, however I'm looking at 4 or 8GB flash IDE drives for my desktop and
> development boxes (with /home NFS mounted off the main server elsewhere)
>

What sort of distro. do you get that installs in 32MB (or have I missed the 
point)?

-- 
Geoff

In article <484f813a$1_1@glkas0286.greenlnk.net>,
Geoffrey Clements  wrote:
>"Gordon Henderson" <gordon+usenet@drogon.net> wrote in message 
>news:g2m8mv$2f9m$1@energise.enta.net...
>> In article <2QZrX2MUwoTIFwFs@tynecomp.invalid>,
>
>[snip]
>
>> I build my mini ITX system booting from flash IDE drive (32 or 64MB), but
>> they all run in RAM after that - so the flash disk sits unused 99% of the
>> time, however I'm looking at 4 or 8GB flash IDE drives for my desktop and
>> development boxes (with /home NFS mounted off the main server elsewhere)
>>
>
>What sort of distro. do you get that installs in 32MB (or have I missed the 
>point)?

It's based on Debian, but it could be anything, as it's a very finely
tuned and stripped down installation.

What I do: format the partition on the flash drive, give it a /boot,
/etc and /dev. Put a bzImage (and a memtest86+) in /boot, put a
lilo.conf in /etc and enough of /dev in /dev.

Then I create a file on my development system the same size as the target
ramdisk, mkfs it and loopback mount it. Copy into that just what I need
in my system, so I just take what I need out of /bin, /etc, /usr/bin,
/lib, etc. to build up my own system. Unmount it, gzip it and put it
on the flash device as /image.gz. The kernel is set to boot, and load in
a ramdisk and it uncompresses image.gz into the ramdisk, and off it
goes... System runs entirely from RAM. (and cd / ; rm -rf * is something
you get bored of doing quite quickly ;-)

This is what it looks like when running:

  Filesystem            Size  Used Avail Use% Mounted on
  /dev/ram0             136M  105M   32M  77% /
  tmpfs                 244M     0  244M   0% /dev/shm
  /dev/hda3              64M  1.3M   63M   3% /data
  /dev/hda1              49M   42M  6.9M  86% /flash

The /data is a separate partition formatted as ext2 - this is force
fsck'd at boot time (if it's not clean) - in this application, it's
used mostly for voicemail storage. /dev/hda2 is a small partition I
store config data into which is unpacked very early in the boot
sequence. (eg. stuff like /etc/network/interfaces, resolv.conf, and
other stuff like that)

/flash is the flash device which I've mounted here - it's not normally
mounted at all unless I'm re-flashing the devices.

  # ls -l /flash
  total 41019
  drwxr-xr-x   2 root root     1024 May  4 19:51 boot
  drwxr-xr-x  13 root root    24576 May  4 11:11 dev
  drwxr-xr-x   2 root root     1024 May  4 11:17 etc
  -rw-r--r--   1 root root 41797882 May  4 11:17 image.gz
  drwx------   2 root root    12288 May  4 11:17 lost+found

My 'base' system + kernel, etc. will fit well inside a 32MB device (not
that I can get them anymore!), but perl & asterisk needs a bit more
(the bulk being the asterisk voice prompts!)

So the down-side is that I can't use the distributions update mechanisms -
I have to update the development server, then re-build the image and
install it on the target box, but it all then runs in RAM which reduces
the "wear and tear" on the flash devices. (and it's faster) For the
PBXs, I used to put in 2 x 64MB devices, one for voicemail, the other
for the system but I now just use one larger device (128MB or 256MB),
with a separate 2 or 4GB device for call recording (an optional extra!)

I haven't been particularly careful about what I include in it - eg. I'm
not using busybox, or ucLib and so on to try to really squeeze it down -
with these systems I can afford a bit of space vs. the time to develop
it. Memory's cheap, and I don't mind if it take a minute or 2 to boot.

I've used the "base" system to build routers, NAS boxes and PBXs. (The
NAS boxes having 2 real spinny things, of-course - and as the motherboard
I'm using has 2 x SATA ports, as well as the IDE ports, they're ideal
for the purpose - 1TB of storage on 2 mirrored drives in under 35 watts...

Gordon

"Gordon Henderson" <gordon+usenet@drogon.net> wrote in message 
news:g2o4qk$2p53$1@energise.enta.net...

[snip interesting stuff]

Thanks Gordon that _is_ interesting. It sounds like you have the start of 
your own distribution there!

I'm thinking of getting one of these cases for a home server, 
http://www.mini-itx.com/store/?c=42, to replace my poorly Pentium II box 
(memtest finds lots of errors).  It has an internel drive bay and 4 removal 
bays and I'm thinking of using a flash drive for the internal one to hold 
the OS and all data will be held on the removable drives (probably mirrored 
but that'll come later).

Cheers,
-- 
Geoff

In article ,
Geoffrey Clements  wrote:
>"Gordon Henderson" <gordon+usenet@drogon.net> wrote in message 
>news:g2o4qk$2p53$1@energise.enta.net...
>
>[snip interesting stuff]
>
>Thanks Gordon that _is_ interesting. It sounds like you have the start of 
>your own distribution there!

I did consider it (rolling a "proper" distribution) - but since I've used
Debian since the year dot, I just went with what I knew. It is a bit of
what might be considered a "gross hack", as all I'm literally doing is
simply using the binaries and configs that that I need, taking enough
of /etc (etc.) to make it work... I just have one big shell-script that
builds it all!

The down-side is that should any gaping security issues come to light,
then it's a PITA to re-flash them all... But with good firewalling, then
I suspect any issues ought to be minimal. This'll be more of an issue if
I stick an ADSL card/usb adapter in them. (See another thread!)

>I'm thinking of getting one of these cases for a home server, 
>http://www.mini-itx.com/store/?c=42, to replace my poorly Pentium II box 
>(memtest finds lots of errors).  It has an internel drive bay and 4 removal 
>bays and I'm thinking of using a flash drive for the internal one to hold 
>the OS and all data will be held on the removable drives (probably mirrored 
>but that'll come later).

Looks like a nice box - if you need the 4 drives... (And if you need
them to be front-panel removable)

I've used this:

  http://www.mini-itx.com/store/?c=3#venus

For SOHO type NAS boxes with 2 drives - WDC ones are favourite right now
as they're very low power (and thus cool!)

I have to say, I'm almost tempted to specify a Drobo for my next client
who wants a NAS type of box - partially biased as I have a friend who
works for them, but even so, it's a neat little box...

Gordon

In message , Geoffrey Clements 
 writes


>I'm thinking of getting one of these cases for a home server,
>http://www.mini-itx.com/store/?c=42, to replace my poorly Pentium II box
>(memtest finds lots of errors).  It has an internel drive bay and 4 removal
>bays and I'm thinking of using a flash drive for the internal one to hold
>the OS and all data will be held on the removable drives (probably mirrored
>but that'll come later).

I was given a mini-ITX motherboard and was looking for a case. I 
eventually tried it in an ATX case, the one that my old P II system came 
from. It fits perfectly. The motherboard and a couple of disks run from 
the old 175w power supply which has a tiny and very quiet fan.



-- 
Bernard Peek
London, UK. DBA, Manager, Trainer & Author.

Gordon Henderson wrote:

[snip]

>>I'm thinking of getting one of these cases for a home server,
>>http://www.mini-itx.com/store/?c=42, to replace my poorly Pentium II box
>>(memtest finds lots of errors).  It has an internel drive bay and 4
>>removal bays and I'm thinking of using a flash drive for the internal one
>>to hold the OS and all data will be held on the removable drives (probably
>>mirrored but that'll come later).
> 
> Looks like a nice box - if you need the 4 drives... (And if you need
> them to be front-panel removable)
> 
> I've used this:
> 
>   http://www.mini-itx.com/store/?c=3#venus
> 

I considered one of those but I'm trying to go completely fanless and they
have a fan in the PSU, however they are much cheaper ... decisions
decisions ...

-- 
Geoff                                           Registered Linux user 196308
Replace bitbucket with geoff to mail me.

The comment on 'wear and tear' on the flash, if you run in this,
rather than RAM, is 'vital'. The write life of flash drives is
suprisingly low, if things are happening at all frequently.
You can get flash 'IDE' drives, designed as replacements for HD's, but
they cost more than the basic memory sticks (not so bad now). On
these, they add extra sectors, and use a randomising algorithm, to
move where repeated writes take place. They also usually add ECC
correction as well. I have used mini distro's, on machines running in
these (and working with them as their 'drives),' OK, but have also
managed to kill a couple of USB flash sticks, trying the same approach
on these...
So though 'running in flash' sounds great, unless you are using a
flash designed to support a high 'write' life, keep to using the
RAM!...

Best Wishes

In article news:<485023c9$0$647$5a6aecb4@news.aaisp.net.uk>, Geoffrey 
Clements wrote:
> >   http://www.mini-itx.com/store/?c=3#venus
> > 
> 
> I considered one of those but I'm trying to go completely fanless and they
> have a fan in the PSU, however they are much cheaper ... decisions
> decisions ...

I have a venus case (not a 669 but the earlier 667(?)) and it's really quite 
noisy for an allegedly quite quiet case.

However ... there are precious few small cases with 2+ 5.25" bays around, 
and the Venus is a nice case ...

There does seem to be an opportunity here for someone to produce a compact 
mini-ITX case with a stack of 5.25" bays and no 3.5" bays that would be 
suitable for use with a small server PC with all drives in removable bays.

I'm considering getting another Venus and discarding its PSU (and case fans) 
in favour of a pico PSU and external mains brick ... but I'd rather have 
something with 3 or more front-accessible drive bays.

Cheers,
 Daniel.