CV11 Server Hard Disk Configuration

This document discusses recommended server configuration for CV11.

Collect! Server Configurations

There are many possible ways of configuring a CV11 Server. This topic briefly describes three configurations -- a typical configuration, a better configuration and then a fast configuration.

Since Collect! is a database application, the typical bottleneck is disk access, therefore the configurations below optimize drive performance.

The ideas below are meant for guidance and discussion only. Professional systems builders should be able to come up with better solutions, but the critical ideas are at least discussed below.

Typical

- Server uses Raid 5 or Raid 10 SCSI drive array for all files.

This is a typical installation on a single server direct from the CV11 installation CD.

Better

- Use Raid 5 or 10 array for Main database files.
- Use separate physical drive(s) and/or raid array(s) for log files.

1. Use the RDS Admin tool to Edit the devices, and change the path to the device 'syslog' to one of the separate physical devices.

2. Change the 'Message file path' field in the Collect! network preferences to another separate physical device.

Fast - Radical

Do the same as in the 'Better' scenario above, but also use the RDS admin tool to split your active database files across more drive(s) and/or array(s) and use a RAM disk for the database log files.

Arrange drives as follows:

1. Rapidly changing database files - Raid array


2. Database log files - RAM Drive


3. Collect! message files - fast drive


4. Index files for the rapidly changing database files - 1 physical
drive per index file?

5. Slowly changing database files and their index files - fast drive


6. Hot online data backups - very large separate drive

For information about which specific CV11 data files belongs in each group above please email support@collectsoftware.cc.

At some point in the early morning, a Script file should put the RDM Server into hot online backup mode and only changed files from all drives are copied to the hot online snapshot drive. Then the server is brought out of hot online backup mode. Later another machine can backup the data from the snapshot drive. A separate NIC from the server to an external tape backup machine would allow a tape backup to run without impacting the performance of the main network.

Technical Notes

1. Physically when Collect! writes a record it not only writes to the file associated with that record but it also writes to index files, and on a heavily indexed record (like the contact record) it may write to 10 additional key files every time a contact record is changed. That is lot of work for a physical disk. We tested upgrading a 1/2 million account 5GB database with indexes enabled and it took over 48 hours but when we disabled the indexes the database upgrade took less than 1 hour. Afterward the indexes needed to be rebuilt which took 8 hours but the whole process was still much faster. Wow! That was on a 500MHz PII machine using a dual SCSI II Raid 0 array of 2 drives.

2. Consider optimizing the main data drive to accommodate lots of small files, since Collect! creates many fairly small note, Attachment, report and other files.

3. Please check the Installation/Administration Guide for important performance tips for tuning your server.

The complete guides for RDM Server are available for viewing from the Raima web site.

The URL is http://docs.raima.com/rdms/8_3/

See Also

- Installation Topics
- RDM Server Reference Guides