The VMware vCenter server provides a centralized way to control and monitor your virtual infrastructure and version 5.1 offers significant improvements to the statistics subsystem. The vCenter server 5.1 includes enhancements in stored procedures in the database which can reduce the resource overhead.
To take advantage of the new improvements it is important to properly configure the database, especially in regards to statistics data which represents a large portion of the overall database and significantly affects its performance. These are some recommendations which can work to get the most out of the Vcenter 5.1 improvements.
Lay Out Disks for Oracle and SQL Server
Because the vCenter server operations can cause a lot of I/O on the database server, it is important to spread disks evenly. For Oracle you will need to have 7 disks and for the SQL server you must have 5 disks to lay the database files.
Update Index Statistics for Highly Volatile Tables
In the SQL server, index statistics are updated automatically as the default setting. However with larger more volatile tables, the SQL requires that a few thousand or millions of rows be inserted or updated before the statistics are automatically updated which can affect server operations. To avoid any issues, you should manually update the index statistics for highly volatile tables for the best database performance.
Separate Tables and Indexes
You can improve performance for very large databases by moving non-cluster indexes onto their own file groups. You should also separate indexes with Oracle, as it can show significant improvement in database response time. Make sure to back up your database before performing these procedures.
Prevent Statistics from Becoming Stale
Statistics in Oracle need to be up-to-date because stale statistics can have an adverse effect on database response. The default setting for Oracle is to gather statistics automatically, which is sufficient for most database objects at a moderate speed, but data on large tables changes rapidly making the statistics become stale. To prevent this you should gather statistics at regular intervals to ensure that they accurately represent database objects.
Use Parallel Query Processing
With the SQL server you can improve query and index operations through parallel query processing which runs several threads in parallel across multiple CPUs. Parallel execution plans use more than one thread while serial execution plans can only run one thread.
To take advantage of these features you can set the SQL server to max degree of parallelism which determines the computing and thread resources used for parallel query processing. You can also set the cost threshold for parallelism which specifies the threshold at which the SQL server creates and runs parallel plans for queries.
The estimated cost to run a serial plan must be higher than the value set in order for the SQL server to create and run a parallel plan for a query. In order to take advantages of new features and improvements in VMware vCenter 5.1 you will need to do a few revisions to tables and procedures for the statistics subsystem. These changes will make the vCenter more efficient to use.
Author Leo Hart is a cloud computing expert from setting up custom virtual servers for his company Custom Cloud.