Failure rate on older servers

[Ninad_1]

Hi,

I wish to know if there are any standard rules/observations for failure rates of older equipment like servers/storage - to be more specific I wish to know if there are any trends of higher failure rates related to older servers , like V2500 and storage array like XP256.
This is required to suggest to the customer to upgrade to latest storage/servers as XP256 is nearing it End of Support in 2007 and V2500 has current EOL of Sep 2006 with ECA till 2009. Thus I wish to highlight if there are any risks in continuing with older equipment and need to quantify the risks in terms of say outage period etc.
Also since HP-UX 11.00 has End of support by Dec 2006, I need to know how to hightlight the risks in continuing with unsupported OS version and better if I can point out to specific problems/am able to quantify the risk/loss to make them understand.

Thanks,
Ninad

[Bill Hassell]

The failure rates on most HP 9000 servers (like the K-class and D-class) are extremely low. However, the V-class was designed by a company that HP aquired several years ago and did not seem to have the same long life as HP-designed servers. Additionally, the V-class consumes enormous amounts of electricity and air conditioning and hardware support costs are also quite high (compared to current servers).

Using unsupported OS software is not only risky dur to lack of fixes (patches), but may cause audit failures for things like HIPPA, SOX and other government regulations. In some cases, auditing failures may require shutdown of the system as there may be no fix possible.

As far as the XP256, these should be fairly stable with the vast majority of failures now limited just to disks.

[Ninad_1]

Bill,

Thanks.
My customer is asking for some statistics/figures/explicit statements saying/showing how failure rate figures look like as the equipment gets older.
Also what are the problems are expected for hardware and OS (need to give some specific type of errors/problems and how they will affect - in the sense the system cannot be recovered/ just needs reboot etc etc)
Can anyone throw more light on this please.

Thanks
Ninad

[Victor BERRIDGE · Accepted solution]

Hi Nidad,

I agree with Bill, it is difficult to give rules/observation on failure rates especially when talking of "High End" gear of the time...
You have invested a great amount to get that "state of the art" equipement at the time e.g. because of its failure rates (XP256 was given for 99.999 uptime) that has no reason to change except the age now will/may bring it to fail one day but the way it is built (fault tolerant...) if failure happens will be transparent for the End-user, not for the people in charge of paying the maintenance!
We just reneewed our HDS9980V (HP-XP512) for the latest model, to justify the expence was the price of the maintenance (End of guaranty...) the new price of the maintenance for 3 years equaled the price of the new system...

Now the risk of continuing with unsupported OS is more why arent you migrating?
And the most obvious answer would be the application has to be rewritten, so the risk is more do you have enough spare to cope with failures till the application has been rewritten or migrated to a new software I still maintain 3 servers in 10.20 (oracle 7.2.3 with forms 3...) 2 are D classes and 1 is a K but I have 2Ds and a K as spares...


All the best
Victor

[Alexander M. Ermes]

Hi there.
We use a V-class since 1998. Downtime rate was mostly once year for memory and cpu failure. But the tco is very high as mentioned before. Eating power like popcorn,
request large amounts of air condition and so on. We use it as a test system.
If you use it in production, replace it asap.
Run in parallel to the new system to test the replacement very carefully.
Be careful, if you use cluster technology from HP with Oracle 9.2. You cannot run sql scripts from cluster disks.
Just my 2 ct.
Alexander M. Ermes

[Alzhy]

Even with the absence of hard stats and just basing this onthe fact that servers do have mechanical parts and temperature variations and materials fatigue are also factors - I would say that failure rates for older equipment - storage and servers alike *should* have higher failure rates.

However, after stating that - in our own data center - I have not seen a specific pattern pointing to this fact.

[Ninad_1]

Thanks a lot for all your inputs.

Alexander,

Why do you say that "if you use cluster technology from HP with Oracle 9.2. You cannot run sql scripts from cluster disks.
" - ? whats the constraint ?

Thanks
Ninad