Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I have a HP server with windows server 2003 and 50 windows XP clients.

Since a week and a half the networks speed suddenly drop 2-3 times per day. It gets so slow that none of the clients could work with the HIS program installed on them.

We tried so many different things such as replacing the hubs,switches and even some wires. Every time one of these changes solves the problem and the network goes back to its normal state.

I checked everything. Even when I disconnected all the clients from the server and connected it to just one computer the problem still remained for 2 hours.

I just narrowed down the problem to the couple of likely speculations as follows:

  1. viruses? (Updated Kaspersky running on the server shows none)
  2. server hardware failure?
  3. Physical memory usage on the server? (Because the last time the problem occurred none of the changes above solved the issue so I restarted the server an checked the physical memory usage which was 2 GBs. But I noticed it's increasing over time to over 9 GBs...the server has 16 GBs of RAM.)

I surfed the internet and got nothing. Any help would do us a lot....thanks in advance

share|improve this question
    
Is there anything in the event log on the server? –  Hennes Sep 9 '12 at 11:05
    
sorry Hennes but how should i get help from the event log?what should i look for? –  farzinSH Sep 9 '12 at 11:23
    
Ok. 0) replacing wires/hubs etc all might have reset the network state. Maybe that is all that is required to temporarily fix it. 1) Virusses are possible, but servers rarely get infected because they are not normally used to run semi random programs. Just the needed and trusted stuff. 2) HW fail is possible but usually leaves traces in the log files (look for red marked entres under system). 3) RAM usage is not a problem. It is there to get used. Unused RAM might as well be in a drawer as in a server. –  Hennes Sep 9 '12 at 11:23
    
Yep...resetting the network temporary fixes the problem but i couldnt find the cause...Any ideas??? –  farzinSH Sep 9 '12 at 11:31

2 Answers 2

Ultimately, you need to know what the server is doing when the problem is apparent, and what the server is doing when all is OK.

Other than saying the network is going slow, you don't really give any clues as to how the server is responding.

When you say that the network is going slow, do you mean that the client application is running slowly, talking to the server, or do you really mean that packet responses are taking a long time?

My action plan would be:

  1. Confirm that the server's hardware is OK. If you have the ProLiant Support Pack (PSP) installed, browse to https://yourservername:2381 (you have this installed, right?). Logon using an account with admin privs and check the state of your hardware
  2. Check the configuration of your network card(s). See what sort of round-trip latency you have between server/client (from the client, run ping -t servername).
  3. Check the event logs (particularly the system log) - run eventvwr.exe
  4. Check the free-space on your drives. Find out where your paging file(s) are, and how big they are. Consider de-fragmenting your drives.
  5. Use performance monitor to examine:

a) Physical disk - average disk queue length (want this to be <= 2)

b) Physical disk - % disk time (don't want this to be constantly > 80%)

c) Processor - % processor time (don't want this to be constantly > 80%)

d) Network interface - bytes sent/sec

e) Network interface - bytes recieved/sec

f) Memory - page reads/sec

6) Finally, a bit lower level, but Systems Internals' (now Microsoft's) Process Monitor and Process Explorer tools are fantastic at providing an insight into what's actuall happening on a server

--- 10/09/2012

So, the server is otherwise healthy and responding OK (can remote desktop to it and "use it"). The reason for stressing this point is that Windows Server doesn't handle kernel resource starvation too well. Linux starts killing processes when the kernel is threatened, but MS haven't cottoned onto this idea yet. When kernel resources are maxed out (non-paged pools, Etc), Windows servers can just stop responding... until something frees up a kernel memory resource. Not having a sufficiently large paging file (or files) can expedite resource starvation on busy servers. My next steps would be:

  • Check SQL counters in perfmon (as Hennes suggests)
  • See how SQL is responding on the server when you are experiencing problems (can you perform basic queries from SQL Management studio?)
  • Check the configuration of your SQL server (memory and CPU parameters)
share|improve this answer
    
thanks a lot Simon for your detailed answer...means a lot to me. to when i say slow i mean nearly frozen cause the HIS applications which are running on clients wouldnt get any responses to their sql requests from the server.its worth mentioning that at the time the problem occurs I checked the remote connection to the server and the ping from the clients and they are ok.besides i ve checked all the things u said except for number 6... by the way i couldnt find any unusual event happening on the server at the time of the problem and in normal network state...why is it happening priodically? –  farzinSH Sep 10 '12 at 7:34
    
SQL requests... Depending on your SQL server their might be a way to monitor if it is this program which causes the delays. Maybe some transaction which get stuck and which get rolled back when the link to a client is lost (e.g. when you reset the network). –  Hennes Sep 10 '12 at 11:44
    
so you are suggesting that it might be caused by the clients sql requests?but The server's cpu process remains 0% at these times and it shows no unusual processing... –  farzinSH Sep 10 '12 at 15:34
    
I've appended to my answer above, as comments are limited in length... –  Simon Catlin Sep 10 '12 at 20:11
    
the thing is that i have another application running on some clients and its constantly sending requests and recieving updates from a SQL table every 10 to 15 secs and this application works fine even at the time the problem occures...could it still be sql? –  farzinSH Sep 10 '12 at 21:15

Problem Solved! It was coming from the unrestricted size of our sql database log file.Our server has a 2 TB of HDD drive and by default sql sets the AutoGrow option for any created DB to 10% of the HDD capacity.In this case our log file went to be 200 GB and that was the reason the network speed dropped every 6 or 7 ours(SQL error 5144).why?Because every time the clients request a query and a change to the DB is going to be logged by the sql server it has to write in the log file and put zero for the rest of the log file size and this causes the problem.To solve this we restricted the size to 10 MB for the log file and Problem Solved... thanks a lot to Simon and Hennes for their reply.

share|improve this answer
    
Also running a transaction log backup at least daily will keep the log from growing, as well as setting the recovery mode on the database to 'simple' instead of 'full'. –  KJ-SRS Sep 14 '12 at 20:32

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.