No announcement yet.

Performance Problems

  • Filter
  • Time
  • Show
Clear All
new posts

  • Performance Problems


    I'm having some problems with a system that i'm checking.

    The users are reporting "freezes" or slow response ( when they type, the letters are shown letter after letter slowly ) while they are working on the terminal servers.

    I didn't found any major problems with the terminal servers and i'm thinking the problems are caused by another server.

    They have 1 server that is used as the file server, and also as their database server.
    the database engine is pervasive 8.

    Memory and CPU usage is fine, ( the server has free resources ) so it's not a problem of weak machine.

    when I'm looking at the performance monitor, I'm seeing massive amount of pages/sec, and also in the "avg. disk queue length "
    the machine has 6 HD, and they are divided into 3 sets of mirror.

    the third set ( drive E: ) is holding the databases, the client software ( magic 9.4 and 9.3 ), and also some small paychek software.

    most of the times whenever there is a peek in the "avg. disk queue length " it is caused by drive E:

    I'm thinking that this is all wrong, and the databases should have their own server, but having some problems convincing the company to build another server.

    what other counters I should check to prove that I'm right?



  • #2
    Re: Performance Problems

    Just a clue as to the operating system and service pack would be really useful
    As would the hardware specs....
    Tom Jones
    MCT, MCSE (2000:Security & 2003), MCSA:Security & Messaging, MCDBA, MCDST, MCITP(EA, EMA, SA, EDA, ES, CS), MCTS, MCP, Sec+
    IT Trainer / Consultant
    Ossian Ltd

    ** Remember to give credit where credit is due and leave reputation points where appropriate **


    • #3
      Re: Performance Problems

      where is your paging file stored ?

      Avg DIsk Queue length represents the amount of time a command spends in a queue before it's processed, which to me suggests a disk i/o issue.. disks might be too slow,srever might just be overloaded and paging like a bitch..
      Please do show your appreciation to those who assist you by leaving Rep Point


      • #4
        Re: Performance Problems

        Sorry for all the details I've forgot to mention.

        operation system : server 2003 Std 32Bit SP2
        HP DL380 G4
        CPU : Xeon 3.2Ghz
        Memory : 4GB
        HD: 2X 72GB 15K SCSI, 4X 146GB 15K SCSI

        The paging file is located on drive C:

        disk are SCSI 15K rpm.

        I've also looked at disk write queue length of drive E:, and it shows a lot of spikes.


        • #5
          Re: Performance Problems

          You haven't told us what the avg. disk queue length is spiking to. What, how long, and how often?

          In my experience with our TS farm a delay in seeing typed characters on the screen and sessions freezing has always been a network problem. Have you looked at the network?


          • #6
            Re: Performance Problems

            does this happen all the time, or randomly and intermittently.. can you pinpoint any processes that might be happening when this occurs ?
            maybe someone is running a large stored procedure, or processing a report ?
            Please do show your appreciation to those who assist you by leaving Rep Point


            • #7
              Re: Performance Problems

              The Problems are not occurring all the time

              avg. disk queue length is spiking to about 4.8 in 100.000 scale

              The network is fine, no problems
              Last edited by Yaniv Hoobian; 7th June 2009, 11:54.


              • #8
                Re: Performance Problems

                The avg. disk queue length spiking to 4.8 is not a problem in my opinion.

                The general consensus on the avg. disk queue length metric is that the queue should not exceed 2 per spindle for long periods of time. So for your first array with 2 disks the queue should not exceed 4 for long periods of time, but an occassional spike above 4 is OK. For the array with 4 disks the queue should not exceed 8 for long periods of time, but an occassional spike above 8 is OK.

                When you say the network is OK, how do you know? What did you look at? Have you run a packet capture?

                You should run a network capture on the TS and look for a large number of broadcasts, ARP requests, duplicate acknowledgements, resets, etc. If you see any of these in large numbers then you do have a network problem. You should also look at the counters on the switch port that the TS is connected to and look for errors. If you have a lot of errors on the switch port then that indicates a network problem. You should also look at the network counters in perfmon and look for errors and look at the queue counter. If there are errors or if the queue is above 0 then that also indicates a network problem.


                • #9
                  Re: Performance Problems

                  we scanned the network with a sniffer, and found nothing but regular communication.

                  I'll run the sniffer again and also the network counters.

                  maybe i didn't explain myself correctly, there are 3 arrays, each one of them contains 2 disks.

                  So you don't think the problems may be generated from the file /DB server?



                  • #10
                    Re: Performance Problems

                    It could be but as I said, in my experience those types of problems have turned out to be network issues. The avg. disk queue spiking above 4 is not an issue IMHO unless it happens very often and is sustained for a long period of time. How often does it spike above 4 and for how long?


                    • #11
                      Re: Performance Problems

                      It spikes during load time, the intervals are about 1 second between spike to spike.

                      And it is caused mainly by writing to the DB array.


                      • #12
                        Re: Performance Problems

                        So it spikes above 4 every other second or so? For how long? One minute, one hour, etc?

                        Are the databases and logs on the same array?

                        How are the arrays laid out? Where is the OS? Where are the databases? Where are the log files? Where is the swap file?


                        • #13
                          Re: Performance Problems

                          3 arrays

                          the OS is on the first array ( C: )

                          shared files are on the second array ( D

                          DB are on the third array ( E: )

                          the log files are on the DB array.

                          Swap file is on the first array

                          today the avg disk read queue on drive E: went to value of 12 for about 3 minutes, several times

                          I notice something else, on one of the terminal servers , a user used a program that cause the pages /sec to about 6500 for 5 minuets.

                          is that something that is acceptable?

                          Last edited by Yaniv Hoobian; 10th June 2009, 09:29.


                          • #14
                            Re: Performance Problems

                            And the arrays are seperate physical arrays of 2 disks each, right? So you have three seperate RAID1 arrays as C, D, and E?

                            The databases and logs being on the same drive could cause performance issues as logs are read/written sequentially and databases are read/written randomly so these two different kinds of I/O contend with each other for the disk.

                            The high number of pages/sec on the TS could indicate a shortage of memory or it could indicate a high amount of disk activity. What is the program on the TS and what is it doing when you get the spike in pages/sec?

                            Have you monitored any other counters on the TS?


                            • #15
                              Re: Performance Problems

                              yes, 6 disks divided into 3 arrays.

                              on the TS was at least 1GB RAM free
                              the program is a paycheck program, I don't know exactly what it was doing, but I suppose it was producing paychecks. it was also writing to the disk

                              I'm monitoring disk, network, memory CPU, counters.