Dean Kent (dkent@realworldtech.com) on 5/23/06 wrote:
---------------------------
>William Campbell (wcampbell@realworldtech.com) on 5/23/06 wrote:
>---------------------------
>>
>>In the real world, how likely is that to happen though. My experience is that
>>once a production platform becomes stable, the people who look after infrastructure
>>are _loathe_ to change to (for example) a new JVM etc. because the validation and
>>regression testing costs are prohibitive. Perhaps I've only worked in overly conservative
>>organisations? So while the theoretical numbers are impressive, whether Woodcrest
>>will be as dominant may not be so clear cut?
>>
>


>Isn't it more of a "cost vs benefit" issue? If there is a significant enough benefit,
>then the cost is justified? Otherwise, there would never be anyone going to new
>releases.

Not necessarily. Platforms become EOL. For example, Windows NT, WebSphere Application Server (WAS) 4.0. My current employer only moves (moved) because it has (had) to. They will (have) moved to new platforms, with the benefits they might bring, but performance concerns don't appear on the radar.

Plus, there are patches that are applied all the time to fix problems
>- and sometimes fix performance problems.
>

Yes, but patches are applied only as a last resort. They consider it more cost effective for applications (application coders) to code around bugs. We operate a number of application servers which are hosted on the same box. If an application on a machine requires patching, all the applications need to be certified for the new patch level. The man hours involved in such testing make the cost of upgrading patch levels exhorbitant, and hence, patches are only applied as a last resort. To patch for performance is anathema.

There exists an extensive review with regards to consolidation of the mid-range server platform. If this goes ahead, this will only make what I described above even more true. Virtualisation holds some hope here, but virtualisation compounds disaster recovery efforts. To be quick and effective, you want the bare minimum of configurations across your deployed assets. If machine A has two virtual servers with differing configurations, then you have to do twice the work when disaster strikes, and you really don't want that. I don't think it is a clear cut issue.

>I think the question is whether any performance bottlenecks can be addressed via
>patches, and how much performance can be gained that way - and how much performance
>could be gained by redesigning/recompiling an application. Would you agree with this?
>

As above. Performance gains are nice, but they are not the movers. At least with my current experience.

>Regards,
>Dean
>