Performance PS4.0

Hi

I have 8 citrix servers with Windows 2003 SP1 and PS4.0.
I have 4 GB RAM and my pagefile is on a different partition with a space of 8GB.

Hardware: HP blade 20p G2

The performance is really slow. Users are complaining about their applications that sometimes just hangs for several seconds before it continious.
When i start performance monitor and trigger the Memory Pages/sec then it goes often to 100 %, even when there is enough RAM available !

What can i do ?

Tahnks,

Robin
LVL 1
rudi_van_de_veldeAsked:
Who is Participating?
 
ET0000Commented:
Some applications are memory hogs - if these apps are CAD or use large databases, there may be that much paging going on. Keep using PerfMon to investigate. Try to determine whether specific apps are causing the paging, and if that ties to specific user activity. Also, look at other variables -- paging might be a symptom of another problem, so keep a watch on CPU, network bandwidth, disk I/O, etc. while monitoring paging.
0
 
BLipmanCommented:
This is a good link for performance tuning and troubleshooting load issues:

http://www.brianmadden.com/content/content.asp?ID=76

What applications are being hosted?  How many users per server?  
0
 
rudi_van_de_veldeAuthor Commented:
Hi

We publish the desktop and we use roaming profiles. But i do not think this is the problem.
I have 290 users on 8 servers.
Most of them are using a huge java application (j2ee) and use 150 MB memory per java.exe process.
So i think i should get more servers but its weird that they complain now and not three months ago.

So i have per server two 72 GB disks RAID-1 and have three partitions on these disks:

C-drive
Pagefile partition
R-drive , this is for the java application to store jar files and things like that for tha application.

Robin
0
Ultimate Tool Kit for Technology Solution Provider

Broken down into practical pointers and step-by-step instructions, the IT Service Excellence Tool Kit delivers expert advice for technology solution providers. Get your free copy now.

 
gsgiCommented:
290 users / 8 = 36 users per server * 150 megs per Java App would require over 5 megs of memory.

I'd see how many java processes are running when they start to complain.
Would it be safe to assume that all 290 users are on every day (nobody is on vacation) and they are all using this java app most of the time?

Are you running regular windows server, or advanced windows server?  Advanced windows server gives access to more than 4gb of ram.
On regular windows server there is 2gb worth of user space for apps, so if your server needs 5.5 gbs of ram than 3.5 gbs of ram are
being paged.  That is a lot.

You need 43.5 gbs of memory to handle the potential memory load (290 users * 150 megs of memory) so I'd recommend going to windows advanced server or windows datacenter server and putting 8 gbs of memory in each server.   This gives each machine 2gbs of OS memory and 6 gbs of user memory, where they need by the calculation at the top a little over 5 megs of user memory each.

Now, this would be the theory.  Hopefully others here will comment on whether or not I am correct.

You could also get a different java app or have someone rewrite it to be less of a memory hog.

-gsgi




0
 
ET0000Commented:
If things were working well before but not now, then something changed, and it might be useful to identify some possible causes. Has the load changed significantly (e.g. more users, cyclical usage like end-of-fiscal-yer deadlines, etc.)? Were there updates applied to the O/S or the application? Have there been any configuration changes?

Does the application reference external resources (a common file, an Internet site, etc.) that might not be responding promptly anymore? (I've seen cases where an application keeps appending to a file, like a logfile, and uses methods that slow down as the file gets larger. That can result in gradual slowdowns that eventually get reported by users more frequently as time goes on.)

Are the complaints coming from the same people, or from people on the same server, or from people doing the same operation? Are there any other similarities that might help focus the investigation?

NB: gsgi's calculations would result in 5 GIG of memory, not MEG.
0
 
ET0000Commented:
Recommend split between ET0000, BLipman , and gsgi.
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.