OE RDBMS slower in the new server than in the old one

Posted by Kai Lahtinen on 17-Oct-2017 02:12

Hi!

One of our customers is facing serious performance issues as they are testing a new production server. The current server is Windows 2003 Server and the new server is Windows 2012 R2 Server. Both are blade servers (not virtual). OpenEdge version is 10.2B08 32 bit. The database is copied using backup files and all the database startup parameters are the very same. The database interbal size is over 300 GB.

When running online backups and batch jobs it can last over two times more with the new server altough during these tests AI was not enabled in the new one. Both servers have SAN disk system and the database is placed on E: (Data), F: (Before Image) and G: (After Image). The SAN disk system is different:

  • Current server, Primary storage platform HDS Virtual Storage Platform (VSP), write cache enabled
  • New server, Primary storage platform IBM SAN Volume Controller IBM_2145:SVC (code_level 7.6.1.7, build 125.2.1702091228000), no write cache

All the data is placed  on type II areas and indexes are on different areas. The total number of areas is 46. Database startup parameters:

  • Database Blocksize (-blocksize): 8192
  • Number of Database Buffers (-B): 80000 (too small because of the 32 bit version)
  • Number of Alternate Database Buffers (-B2): 100
  • Current Spin Lock Tries (-spin): 50000
  • Before-Image Cluster Size: 524288
  • Before-Image Block Size: 8192
  • After-Image Block Size: 8192
  • Number of After-Image Buffers (-aibufs): 40

BIW, AIW and APW are used.

Any ideas what might be the problem? Our plan B is to update to version 11.7 64 bit ASAP.

Regards

Kai

OpenEdge RDBMS - Forum

All Replies

Posted by George Potemkin on 17-Oct-2017 02:24

> When running online backups and batch jobs it can last over two times more with the new server

Can you test online backup to NUL? If it's slow then the reads from disks are slow.

Posted by ChUIMonster on 17-Oct-2017 05:43

Your old SAN has a write cache and the new one does not?  That's a fairly significant difference...

What are the differences in CPU configuration between old & new?  That is also important.

Posted by cjbrandt on 17-Oct-2017 07:46

If reads from memory are slower, make sure the new server isn't using NUMA architecture.

Use Windows perfmon to look at disk queue and IO rates on the different servers during a database backup or a database analysis.

Posted by DimitriG4 on 20-Oct-2017 16:52

Write cache should be enabled  temporarily on the SAN (at least for testing) to give the new setup a chance to compete with the original setup before chasing the people selecting the new hardware out of the building.

This thread is closed