Hi, Our SharePoint 2007 implementation supports 8000 users. Recently we have started seeing issues in performance, sluggish behaviour and users getting "out of memory" exceptions. The servers have plenty of memory, and the application pool max memory never seem to be getting fully utilised. I expect the performance issues are due to the network. Downloading a file from the server rarely reaches 30Kb/s, and if hundreds of users are uploading/downloading ~2MB files, will we have a treading issue? E.g. All of the threads are occupied (and possibly locking resources) while sending content over the slow network. I need some proof to bash the network chaps with. Does anyone know how to get proof that the network is a bottleneck? Are there counters in performance monitor that I can use? Please help. Gary.

Hi Gary,There are several reasons why you would get poor performance issues on a Sharepoint farm.Without further information on your setup, I'd advice you to find whatever best practice is available for SQL performance and SAN setup, for Sharepoint, to make sure all requirements are fulfilled there.By the sound of your implementation, it sounds like a fairly large company and I suspect the network being one of the parts that should NOT be a problem unless you experience congestion on other servers aswell.George Lazo

Have you monitorized the servers with performance counters?What can you infer by looking at the Diagnostic Logs?Can you describe your setup? #of servers, RAM, OS (32/64b), RAM in the Sql Server.Have you developed any custom webpart? Have you tried passing SPDisposeCheck over your assemblies?At first the out of memory exceptions don't have many relations with the network performance...Hope it helpsEnrique Blanco PMP, Consultant www.eblanco.com

Hi Gazz1e,there are many reasons may be caused this performance issue, and you can monitor some performance counter for identfying tha cause of this issue, so i advice you apply stress testing on your farm , this test is a good way for you.you can download the tool from http://sptdatapop.codeplex.com/releases/view/1214the following links may be helpful for you:http://msdn.microsoft.com/en-us/library/ee413885.aspxhttp://blogs.msdn.com/sanjaynarang/archive/2009/04/18/sharepoint-performance-stress-load-testing.aspxhttp://www.helloitsliam.com/archive/2008/03/10/moss2007-testing-using-visual-studio-team-system-2008.aspxBest Regrads, Ahmed Madany

Thanks for all the replies. I have checked the CPU, Memory, Disk I/O on the 2 front end web servers, and the database server. The web servers are 32 bit and never have less than 1.5Mb of available memory. Average CPU activity rarely reaches above 40%.The app pools are configured to use 1.5GB. They rarely use more than 1Gb of memory. The service is hosted on a Bladeframe with four 1Gb/s network cards. The entire network to the bladeframe is throttled to 16Mb/s. This bladeframe also provides antivirus updates to 10000 users. Any remaining network capacity (6Mb/s?) is shared between 4000 to 8000 Sharepoint users. We primarily use SharePoint for content management. Complete and utter madness I know.Fortunately, the database server is hosted on the same bladeframe, so traffic between the web servers and DB don't go through the throttle.The idiots who manage the network are refusing to acknoledge that the network is an issue. I need some proof that the network is causing performance issues on the SharePoint application. I was hoping that I could use performance monitor to detect the total number of threads waiting for the network. Is this possible? Or can I get any other evidence that I can bash them with?Alternatively, if someone can give me a job in the UK, this will also resolve the problem (as it won't be my problem anymore).

Without diving too deep into technical details there are some granular issues you must adress first.Is the pagefile separated from the OS disk (not only partition but physical disk separation aswell, which can be an issue in smaller blades with only two disks and a mirrored setup)?Are the indexfiles placed on a SAN spread over 6 spindles or more (this can be an issue when speaking to storage people who consider it being a load of cr*p)?Are tempdb files setup properly on the SQL as to best practice (500mb filesize, ammount of files roughly same as ammount of cores up to atleast 10 files)?As you probably understand there are numerous things you have to make sure is optimized for the application, before looking outside of the box itself.If the blades reside in the same enclosure, your throughput will be well over the throttled network that you express so this shouldn't be an issue internally between the systems, it can however impact the user experience if the outgoing NICs from the enclosure is suffering from heavy load.I can't offer you a job in the UK but in Sweden it's currently mayhem when it comes to finding good infrastructure/architect tech's for Sharepoint.George Lazo

Ahmed, This information was really very helpful, thanks for sharing. http://support.jesoba.com Errors, Troubleshooting, Problem Cases, Performance Monitoring, Best Practices etc.