Memory Pages/sec ?

[TheMac28nearly]

We are running 2 wsk sp2/ metafrmament 1.8 sp3 servers, each with 2gb mem and dual PIII 800. WE are running load balancing fir 90 users with office and other apps installed

We constantly get peaks of 180+ on this. What is the benchmark for memory pages/sec and are we increasing it ?

Thanks

Andy

 

[CitrixEngineer]

I note you say "Office and other apps".

The "benchmark" is a personal thing really, especially where apps other than "standard" are used (note, there is no such thing as a standard application on a Citrix server).

The ideal would be to build a test server that is identical to your production servers, and monitor it at various stages to get your own reference benchmarks. This would involve a pilot scheme with some of the end users, so that you could correctly gauge how they will use the delivered system.

Your setup looks fine on paper (or screen!), but what apps in particular are paging so frequently?

I hope this helps

 

[TheMac28nearly]

Thanks for the reply, i should have been more specific.

We have 90 users MAX, all capable of running office XP, IE 5.5 sp2, and saleslogix (crm system.) Then we have an accounts package, visio, frontpage, where about 20 of those users have these published. Then some other ones like Pagemaker and Adobe Acrobat Writer which about 5 people have access to

Funnily enough, we have a spare server. We were going to put it into the farm straight away to test the difference, but will re-consider.

Is it possible to nail down performance per app without kicking everyone off and doing a test? Ive done it per user via perfmon.

Thanks for your help

Andy

 

[CitrixEngineer]

I've put up a couple of "FAQs" on test environments on this forum.

I believe that there are issues with Adobe Acrobat Writer - I've seen postings on this forum, but have not implemented this application.

If your test server is built identically to your live servers, and you use load-balanced published applications, there wil be no need to force users off the system; instead, you can force a couple of (well chosen!) users onto the test server simply by denying access to them on the live machines. They need never know that they are using the application on a different server - if you don't want them to - and you can obtain your benchmarks based on individual users, applications or combinations.

I hope this helps