Improving Input/Output Throughput

The distribution of input and output across multiple disks is an important factor in database performance. If consistently high input/output (I/O) occurs on one or more disks housing the database, service time on that disk degrades. To address this problem, replan the database layout to improve application performance.

You can split the OpenText PPM database into the following segments:

  • OpenText PPM tables

  • OpenText PPM indexes

  • Redo logs

  • Rollback tablespaces

  • Temporary tablespaces

  • System tablespace

  • Tablespace for management and related utilities

We recommend that OpenText PPM database instances with moderate transaction volume (instances with more than 5,000 requests per month) have at least four discrete disks, divided as shown in the following table.

Table 7-1. Database disk recommendations

Disk

Recommendations for Data Placement

1

OpenText PPM tables

2

OpenText PPM indexes

3

Redo logs

4

  • Rollback tablespaces

  • Temporary tablespaces

  • System tablespace

  • Tablespace for management and related utilities

For OpenText PPM database instances that have higher transaction volumes (more than 10,000 requests per month), we recommend you do the following:

  • Place each piece of the database on its own separate disk.

  • Stripe the data and index tablespaces across multiple disks to provide adequate disk throughput.

For OpenText PPM database instances with an extremely high transaction volume (over 25,000 requests per month), move specific tables and indexes to separate tablespaces on separate disks. This provides better control and further increases available I/O throughput.