Hello THG experts, here's an interesting problem for you:

We are creating a large-scale interactive art work with a projected 10 year lifespan. The master controller runs on a windows pc. I want to install it and walk away forever and hope that technology and redundancy can help with this. I will be able to remotely login and I can have local staff reboot the machine, but nothing more. The requirements for this box are:

Must be fast
Must be rackmounted
1 GB memory
1 free PCI slot
doesn't need accelerated graphics
disk(s) can be slow, everything is loaded in memory
software is only 200 MB total + OS

Things I know already:

The box will have a UPS
The room/rack is cooled.

The tough parts:

Can one create a system with redundant memory?
How often to drives fail?
How many drives should a RAID 5 array have so that 2 can fail.
Can a mirrored RAID system have more than 1 mirror (eg: 4 drives with the same data)?
Can I create a persistent RAM disk?

Any other suggestions? Any help is much appreciated!

More about the artwork here: <a href="http://electroland.net">It's called "EnterActive</a>."

Damon Seeley

Get someone to run it for you at $X/yr whatever ur budget is. That is as close to perpetual life as u can get.

Nothing can be guaranteed to last 10yrs. Further, the technology will change so much in 10yrs that you will be hopelessly outdated long before then.

Maybe in 1000yrs when things settle down can u expect something to run virtually unattended for 10yrs.

Write an iron clad contract if you want it like that. If you find someone willing to undertake it you can probably sue them into poverty at some time. They might do it buy buying everything in multiple redundancy now. Best thing is to replace some components over time.

The loving are the daring!

10 years without dusting or anything would be very hard. Your best bet would be to use old parts. PII's for example could be passively cooled and were very hard to damage from heat. Those big 5.25" full height Seagate drives might last 10 years, I don't think they rely on air to hold the heads up, but rather use a physical gap on a rigid arm.

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>

Wow, I agree, 10 yrs is a long time. But... Raid5 is only 1 drive with parity data (3 disks, 1 with parity, hypothetically... 12 disks, 1 with parity), so only 1 drive can fail. There is a raid spec (6? I don't remember offhand) that uses 2 drives for parity data. Or you could do what I think is called raid 10, or mirrored raid 5. e.g. 6 disks (or 8, 10, etc.), 2 r5 arrays that are mirrored.

Since you said the whole thing is only 200Mb, you could also possibly use solid-state drives - maybe there's something like a flash drive that can be booted off of. That means no moving parts to fail if combined with a passive-cooled CPU, etc. Only thing you'd need then is redundant power supplies.

Mike.

is this thing going to be on the moon or somthing, lol. they cant expect any computer system to not require hands on maintenance AT LEAST once in a 10 year period.

this is my boomstick!

A 10 year life span on an unmaintained windows machine isn't feasible. Security updates alone are going to cause you to have to do maintenance... otherwise you'll be "compromised" inside 2 years. No questions asked. Lucky if you make it 6mo really.

<A HREF="http://www.simpletech.com/webspeed/industrial/briefs/R1191.pdf" target="_new">http://www.simpletech.com/webspeed/industrial/briefs/R1191.pdf</A>
Check that out, makes it a bit easier, as it has a true ide interface you don't have to worry about custom software... They're not cheap however.

If you're going to use standard hard drives don't use IDE use scsi.

You want raid level 6, but I don't know of a good controller that supports it. Functional storage is (Total Drives-2)*Smallest Drive in Arry. Minimum of 3 drives to implement.

You're going to want a multiprocessor server board of some sort.

My personal suggestion would be re-evaluate the windows requirement and use an off the shelf sun.

I think a better solution than trying to make one system extremely redundant is to make two systems moderately redundant and have it fall over in the event of a complete failure.

Edit: My reasoning for this is that it takes into account memory/motherboard/processor failure... and even better yet if you need it completely redundant, you'd store the backup system at a seperate location incase of fire, flooding, catastrophy.

Eg: 2 Systems Configured as:
Duel Xeon, 2g Mem, 3 SCSI hard drives raid 5'd, redundant power. Sync the data to the backup on an hourly/bi-hourly/realtime. Chances are you will lose at least one of the systems in 10 years, but if its the back up system, you fix it. If its the primary then the backup comes up and takes the place of the primary with only 60m/30m/a few seconds of data loss.

To be completely honest, I wouldn't even consider a project that longterm with this kind of criteria for windows. A good example, that will perhaps bring this to light a bit better:

Windows, 10 years ago. 1995... windows nt 3.5... If there's a windows nt 3.5 system running on the net *still* that isn't a massive admin time sink i'd be mindblown.

I know there are still solaris 2.4 (came out around same time as nt 3.5) machines out there running services for various things (dns, webservers, etc). One isp i worked for had several *ancient* irix and sunos boxes running bind and fall back sendmail servers that had uptimes (aka since last reboot) in excess of 2 years (massive long term power outtages were what caused reboots generally on them.)

Edit2: You guys are doing some astoundingly neat stuff (just looked at site) however, if you're planning on long-term maintenance free projects try to look towards a different platform than windows as a server. Unix, linux, or macintosh are all better long-term options. Windows is a nice enviorment to work on for end users, because most everyone is familar with it, but it's not so hot in the backend of things. Requires many fold the maintenance, many fold the admin hours, and just flat out isn't nearly as stable.

<A HREF="http://www.lochel.com/THGC/php/shadus.php" target="_new">Shadus</A><P ID="edit"><FONT SIZE=-1><EM>Edited by Shadus on 01/16/05 03:50 AM.</EM></FONT></P>