SoylentNews is people
SoylentNews
El Reg reports
A chap named Ross, says he "Just switched off our longest running server".
Ross says the box was "Built and brought into service in early 1997" and has "been running 24/7 for 18 years and 10 months".
"In its day, it was a reasonable machine: 200MHz Pentium, 32MB RAM, 4GB SCSI-2 drive", Ross writes. "And up until recently, it was doing its job fine." Of late, however the "hard drive finally started throwing errors, it was time to retire it before it gave up the ghost!" The drive's a Seagate, for those of looking to avoid drives that can't deliver more than 19 years of error-free operations.
The FreeBSD 2.2.1 box "collected user session (connection) data summaries, held copies of invoices, generated warning messages about data and call usage (rates and actual data against limits), let them do real-time account [inquiries] etc".
[...] All the original code was so tightly bound to the operating system itself, that later versions of the OS would have (and ultimately, did) require substantial rework.
[...] Ross reckons the server lived so long due to "a combination of good quality hardware to start with, conservatively used (not flogging itself to death), a nice environment (temperature around 18C and very stable), nicely conditioned power, no vibration, hardly ever had anyone in the server room".
A fan dedicated to keeping the disk drive cool helped things along, as did regular checks of its filters.
[...] Who made the server? [...] The box was a custom job.
[...] Has one of your servers beaten Ross' long-lived machine?
I'm reminded of the the Novell server that worked flawlessly despite being sealed behind drywall for 4 years.
(Score: 4, Interesting) by Unixnut on Friday January 15 2016, @07:16PM
Those old Sparc machines were built like tanks, I can attest to it. This was while I worked at a very large company that shall not be named (they are well known, and one of the biggest and richest companies on earth), they had an Ultra60 (running some ancient Solaris) which had been there since 1998, running non stop. And it wasn't doing some unimportant stuff, this Ultra60 was the fulcrum upon which the entire firm relied for its functioning. Some of the most important data paths through the company went through this one system (there were a pair of Ultra60's, in case of failover was needed).
Essentially the banks systems grew around the Ultra60, and people just came to rely on it always being up, there and doing its job. After years and years, nobody knew how the Ultra60 worked except one guy who was old school Unixbeard. He basically maintained that machine. When it finally came up to his retirement, upper management collectively shat a brick, worrying that everything was going to die once he walked out the door (they offered him so much money to stay on as a consultant, but he wasn't interested).
After he left, I ended up maintaining the Ultra60 (because I was the only one interested in what the old guy did, and paid attention to what he told me), and it was a magnificent beast, very underpowered compared to the latest machines, but it would sit pegged at 100% for months on end without breaking a sweat. You push one of those new machines to 100% for longer than a day or so they start keeling over, or having a performance drop off (not to mention Linux just kicks the bucket).
When I left the company, the Ultra60 was still going strong, and I don't doubt that deep in the bowels of the building, under the layers of "private cloud", "web 2.0 dynamic", "node.js" and other bloated crap, that machine is till chugging away, making sure work gets done.
Main reason nobody replaced the machine (and why you end up with these ancient systems running forever), is because it ran some custom software that was written decades ago (in perl + C), and
a) the guy who wrote it has long since left and nobody can get in contact with him
b) nobody can read the code, there was no source for the C, and the perl was just line noise
c) It was hacked together in places, and there was no guarantee that someone didn't come to rely on some odd behaviour or some really old function that has been superseded
and most importantly:
d) If you do replace it without a hitch, nobody will notice nor will you get a raise/promotion. However if you mess it up and break it, the whole company stops working, and you will be kicked out of the company so fast you will leave skin marks on the parking lot.
Essentially, it wasn't worth anyone to risk sticking out their neck for it, so it will keep chugging along I guess, until it finally dies (if it does, the result will hit the news, so I will be informed one way or another :) )
(Score: 0) by Anonymous Coward on Saturday January 16 2016, @02:55PM
Sun built solid hardware back in the day. The Sparc based pizza boxes and bread boxes just ran & ran until the SCSI disks died. The Ultra 1 and Ultra 60s were solid too. They tried to cuts costs with the Ultra 5s and 10s. We had some 440 and 880s that lasted too!
(Score: 2) by Unixnut on Sunday January 17 2016, @08:27AM
Indeed they did, and we still had v440's in production. Some teams swore by them for their reliability and sustained performance (especially compared to the x86 alternatives), and were unwilling to let go of the machines. It was only after Oracle took over Sun and hiked the licensing costs by 10 times, that the accounting department and management started pressuring everyone to move off Sparc/Solaris.
The Ultra60 however didn't cost much (if at all), due to being out of support by Sun/Oracle, assuming they even knew of their existence at the company.
(Score: 1) by plnykecky on Sunday January 17 2016, @11:02AM
We have a SPARC 5V working as a Scanning Tunneling Microscope control computer since 1999, but the computer was already refurbished when we got it. It runs Solaris 2 I think and is hooked up by a 10cm-wide ribbon bus to a 8086 processor that collects data in the real time and hands it over to the SPARC. The disk has been replaced some 6 years ago, using dd to clone the contents. I also had to change the SPARC computer at some point and hack its hardware ID to force the old SW license work.
It hangs occasionally but otherwise works fine. We still have the original tapes with the software for it, but no one ever needed a reinstall since 1999.
Next week it is going to be replaced by a control computer running Win7. How sad..