Stratus plus Red Hat AS 4: Lots of nines (NewsForge) [LWN.net]

NewsForge covers Stratus Technologies' latest server offerings, which will run 64-bit Red Hat Enterprise Linux. "While Hewlett-Packard's NonStop Computing division -- formerly known as Tandem Computers -- is a formidable opponent in the world of fault tolerance, Stratus is making inroads in the market with its Intel-based lineup of less expensive, but still highly reliable, servers. How highly reliable? Stratus updates a speedometer at the bottom of its home page daily with a 60-day rolling average of the uptime of its fleet of ftServers around the globe. It shows 99.9997% at present, which equates to about 95 seconds per year."

(Log in to post comments)

Stratus plus Red Hat AS 4: Lots of nines (NewsForge)

Posted Jun 19, 2006 16:06 UTC (Mon) by mattdm (subscriber, #18) [Link]

Note that if you go to stratus.com and click on the meter, it says "The Uptime MeterSM measures the actual availability of installed Stratus® ftServer® systems worldwide running Microsoft® Windows®". (The article doesn't ever say otherwise, but it's not necessarily obvious if you're skimming.)

Plus, while it's an impressive number, remember that it's an *overall* number, and your actual system could be down for much longer than 95 seconds if you're unlucky.

Stratus plus Red Hat AS 4: Lots of nines (NewsForge)

Posted Jun 19, 2006 16:58 UTC (Mon) by azhrei_fje (guest, #26148) [Link]

Actually, if they're running Windows, I'm even more impressed. My wife
can't keep her new Media Center machine up for more than a week or two
without rebooting it. :( And it certainly doesn't reboot in 95 seconds!

Stratus plus Red Hat AS 4: Lots of nines (NewsForge)

Posted Jun 19, 2006 19:29 UTC (Mon) by DG (subscriber, #16978) [Link]

What would be more interesting is how much such protection costs, what infrastructure is required and how easy it is to maintain/manage :-)

Stratus plus Red Hat AS 4: Lots of nines (NewsForge)

Posted Jun 19, 2006 20:44 UTC (Mon) by drag (subscriber, #31333) [Link]

I would expect the answer to the cost of something that highly aviable is:
If you have to ask; you can't afford it.

As for how easy it is to manage:
Doesn't matter. For that price you'll manage irregardless. Mainly because the applications your using it for dictates that you have no choice.

Stratus plus Red Hat AS 4: Lots of nines (NewsForge)

Posted Jun 20, 2006 11:27 UTC (Tue) by markhb (guest, #1003) [Link]

Exactly. According to the article, these boxes appear to be designed for a particular market: telco central offices (the CO being the place where all the phones come together for a particular exchange). These types of machines also show up frequently in the financial industry (the machine the ATM talks to when you ask it for money at 3 Am), and other services that require 24x7 that really means 24x7 (the SABRE / Travelocity database runs on a cluster of 12 HP NonStops; I can't remember where I read that, but Google has plenty of links on the HP / Sabre relationship). I'd be interested in seeing a fault-survival comparison of these Stratus boxes vs. the HP kit.

Stratus plus Red Hat AS 4: Lots of nines (NewsForge)

Posted Jun 21, 2006 7:19 UTC (Wed) by Wol (guest, #4433) [Link]

One of my friends used to work for Stratus (back in the days when their kit was 68000-based).

He told me it was normal for customers to receive spare parts, along with a note saying "your cpu has failed. Please ring and we'll talk you through swapping it over", and this would be the first intimation they had of a problem with the computer!

It was a simple case of "open the case, identify the failed board, pull it out and put the new one in". The computer itself would just carry on running.

Cheers,
Wol

Stratus plus Red Hat AS 4: Lots of nines (NewsForge)

Posted Jun 30, 2006 18:24 UTC (Fri) by djl (guest, #38741) [Link]

The Stratus home page shows either the Windows uptime meter (been shipping about 5 years with several thousand installed worldwide) or their older (20 yrs) VOS operating system - both run about 99.9996 to 99.9998.

What I think is interesting for the Linux community, it this technology provides much higher availability than clustering with none of the complexity. No modification of the app is required