Plex 2E

We had to change an application from WinC-AS400 to WinC/WinNT with SQL Server  and even though we new it was not going to be the same performance the difference is 2 to 1.   The application on WinNT is extremelly low.   We are using the Application Server with the Dispatch manager.   The application has 300 users and we have all the client and server functions on one server and on each user machine we only have the .exe and .ini of the application.   This is with Plex 5.1.Does anyone know any tip on how to configure de Windows Dispatch so that we can improve performance?  Has anyone  compared if  it is better to use  the Plex  Application Server  or just have everything with ODBC?  Thanks for the help and suggestions.      

droca,  Can you tell us how you generated the database on your SQL Server?   Am I safe in assuming it was done from Plex?   If so, how are  your Database Build settings (from Generate and Build Options --> System  Definitions --> Properties --> Database Build) currently configured?   Off the top this sounds like it might be a lack of indexes.    John

Having the client dll's on a server will also cause a huge performance problem as it will be loading the dll's across the network. You actually defeating the purpose of running a client server configuration by doing that.  I agre with John as well that you are missing indexes. On the iSeries this is handled for you, but on SQL you have to specifically say you want an index.

I agree with John and Gavin's suggestions concerning indexes and network deployment.  I will also add that the Position GE statement, as used in BlockFetch's and Get Sequentials, can introduce performance problems when switching from RPG/DDS to SQL-based data access. This should not be a surprise (it's been documented in the Plex help text forever) but still seems to come up as an issue quite often. If you were on an up to date release of Plex I would suggest using StatelessBlockFetch instead of BlockFetch and using C# or Java instead of the old WinNTC generator. Since you are on the old release, consider using Exec SQL to optimize the SQL in BlockFetch functions that are causing performance problems. And to address the original question, if you are running a WinC-WinNTC-ODBC application on a single server then the WinC-WinNTC remote calling mechanism is probably going to be adding unnecessary overhead to those function calls and a simpler WinC-ODBC might be quicker. How much? I'm afraid I have no idea but my gut feel is that it's more likely that data access improvements (Indexes, Exec SQL) will be more significant. Message Edited by DanielLeigh on 02-05-2009 05:39 AM [left]