[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: AlphaServer - Uptime/reliability




On 28 Mar 2008, at 9:51 am, Pearson David wrote:
This is a little bit of an open request but for the installation and
application I'm trying to justify it would really help me if I could get
some 'real' uptime data from the wider industry on how reliable
AlphaServers are. I know they have outstanding performance and are used by half of wall street etc. but it would be great to know from the real
guys who have to keep these things running the real facts behind the
claims.
The two models I'm interested in are Alphaserver DS25 (a number of, both single and dual processor) and an Alphastation DS15. All will be running
Tru64 Unix 5.1b.
Simple feedback such as "Typical application, date installed,
MTBF/uptime/failure rate etc., whether frequently power-cycled or
running 24/7" would be great. Direct contact details to confirm this
would also be appreciated but not essential.
The reason for this request is to support the use of the Alphas in a
mission-critical role running a legacy app as part of the business
justification.
Posting to the forum or mail to hp@pearson.in .

If you want a number of answers to that question, you should ask the Tru64 managers' mailing list, not the Debian mailing list. See http://www.ornl.gov/lists/mailing-lists/

In my experience, Tru64 of itself has not been much more reliable than Linux *except* in the area of high availability clustering tools. TruCluster is more stable than Heartbeat, for example. Our Alphaservers are starting to show hardware failures now, but then they are over 5 years old, many of them considerably more. Some aspects of Tru64 are quite nasty, and unlikely ever to be fixed. For example, the NFS server in Tru64 has some nasty bugs in it, which cause real problems for Linux clients using 2.6.18 or later.

To be honest, I think you need to ask yourself whether investing in Tru64 hardware and software is a sensible bet these days. You can't buy new alphaservers from HP any more (they stopped taking orders last year) and this year they are stopping selling upgrades of any kind. Tru64 development has more or less stopped, and it's increasingly difficult to get open source software to build on it, for one thing because libtool just doesn't work well on it any more. Anyone who's tried compiling glib2 and gtk2 on Tru64 5.1B will know exactly what I mean.

Good quality X86 hardware is just as reliable now as the AlphaServers were. We use a mixture of X86 server hardware from IBM, HP and Sun, and it's all pretty reliable.

Linux is capable of doing 90% of what Tru64 can do, and in some areas rather more than Tru64 can do.

Tru64 is a dying platform; putting money into it now is throwing it down the drain - you're going to have to move to something else -- probably Linux, but this is a Linux list so I would say that -- anyway, so you might as well do it now. You could stick with proprietary UNIX, and go to AIX or HP-UX, but quite frankly I think proprietary UNIX in general is as good as dead. It's just taking IBM and HP a bit longer to realise it than the rest of the world.

The place I work is doing that steadily - we used to have more than 50 ES45 servers, and several hundred smaller DS10's DS20's and the like. We're now down to only 28 Tru64 nodes left in production, and 16 of those are going to be replaced with Debian Linux servers next week. We're trying to get rid of Tru64 as fast as possible now; it's just a management headache now that we have so much stuff on Linux (more than 1,200 servers and 300 desktops running Debian at the moment).

Regards,

Tim


--
The Wellcome Trust Sanger Institute is operated by Genome Research Limited, a charity registered in England with number 1021457 and a company registered in England with number 2742969, whose registered office is 215 Euston Road, London, NW1 2BE.
Reply to: