performance issues - Lawson port to Windows, Infor 10 Upgrade

 5 Replies
 1 Subscribed to this topic
 27 Subscribed to this forum
Sort:
Author
Messages
JimK
Basic Member
Posts: 10
Basic Member

    Hello,

    We are porting our Lawson system from Solaris / DB2 over to Windows / SQL Server and also from Lawson 9 to Infor 10.

     

    We are partnering with a vendor and have completed the first data migration.  Unfortunately, we are seeing much slower performance on the Windows system compared to the old Solaris box.

     

    When we did our homework last fall, we were assured that we could make a Windows system perform as well as Solaris.  We’ve followed the server recommendations and do not see any bottlenecks on the new Windows servers – CPU, RAM, disk, and network all look fine.  We have one batch job that we are using to test performance – the GL290.  It is taking 20 seconds on the Solaris system versus 40 seconds on Windows.  When we run the program with the ‘Timedstats’ utility turned on, we see that calls to one table in particular are taking 12 seconds longer on SQL Server.

     

    We’ve checked the database indexes, and they are all there.  On the old Solaris system, we had the web, app, and database servers all located on the same box.  On Windows each has its own server.  Note that the new Windows servers are all virtual machines.

     

    Has anyone run across this before?  Any tricks on how to resolve it?

     

    Thanks!

    Jim Krzyzewski

    John Henley
    Posts: 3353
      When you were on Solaris was database on same server as the apps?
      On windows is that still the case?
      Without knowing your particulars, I would assume separate servers and would start there.
      Thanks for using the LawsonGuru.com forums!
      John
      JimK
      Basic Member
      Posts: 10
      Basic Member
        Hi John,

        Yep - on Solaris the apps and database were on the same server. Windows system has the apps and database on separate VM servers.

        One of the next things I'm considering is having our DBA create the SQL Server instance on the same VM server as the apps. This mimics more closely what I have with Solaris.

        We don't have a ton of volume - we are an association with about 1,000 employees. Do you think Infor's claim is valid - that a properly tuned Windows system should be able to perform as well as my legacy Solaris system? Note that the Solaris system is relatively new - we purchased new T4 servers for this when we upgraded to 9.0.1 two years ago.

        Thanks!
        EricS
        Veteran Member
        Posts: 80
        Veteran Member
          We see similar behavior on our Windows system. Same set up, multiple VM machines. There is a separate thread (below) that looks very similar. The most odd thing I've found is that the behavior is related to my being forgetful (WHO ME!!!??? at 50 something) and leaving a LID or Portal session up overnight on the LSF server. I try diligently to make sure that all Lawson windows are closed. Any others do not seem to bother it at all.

          https://www.lawsonguru.co...-system-performance/
          John Henley
          Posts: 3353
            You might have to do some tweaking, but Lawson on Windows for a 1000-employee organization is just fine. I'd guess your OS and DB required some parameter tweaking and knob-turning when you started using them on Solaris, right? Shouldn't you expect same thing for Windows? One of the things that clients don't often realize when architecting servers for Lawson S3 (from 7.x thru still to 10) is that the way Lawson's apps use the database is essentially as a storage device, and each record retrieved is essentially a network roundtrip if your apps and DB are on different servers. Throw in how some DBAs layout the databases on a SAN (I'm guessing you probably didn't layer to deal with on Solaris either, correct?) Now, some of that has changed over the years and gotten better, but it can still be an issue. Since you say the app server and database are on VMs (which is fine), I'm assuming they are on same VM host or at least same cluster? You are starting in the correct spot by looking at the timedstats. Now what you need to find out is whether the long waits are due to network latency between the apps server and the DB server, if the wait is the DB server waiting for IO, or if the wait is internal on the DB server.
            Thanks for using the LawsonGuru.com forums!
            John
            John Henley
            Posts: 3353
              Eric's response reminded me to ask if you are using Broadcom NICs on any of the servers. If so, throw them out and get Intel ones. Don't ask.
              Thanks for using the LawsonGuru.com forums!
              John