MQ V9.1 introduces a new HA capability into a long term service release. RDQM (replicated data queue manager) is a high availability solution that is available on Linux platforms, and was first introduced in the V9.0.4 CD release of MQ:
RDQM in MQ Advanced 9.0.4
A performance paper on RDQM HA capabilities is now available on the IBM MQ performance GitHub page: MQ Performance Documents
The paper compares throughput of an RDQM HA environment with a corresponding Multi Instance Queue Manager (MIQM) deployment (using NFS storage), and a non-HA single queue manager, with its transaction logs hosted on a local SAN. Highlights of the report are:
- Over 43,000 round trips/second peak messaging rate in RDQM HA enabled scenario (~86,000 messages produced and ~86,000 messages consumed).
- Peak messaging rate for RDQM HA is faster than MIQM equivalent.
- RDQM HA scenarios with SSD backed DRBD storage, run up to 82-90% of rate achieved by a stand-alone QM, logging to SAN.
- RDQM recovery from fail-over is faster for RDQM than MIQM in all scenarios tested.
Transaction rates of RDQM, utilising a local, replicated filesystem, hosted on SSD’s, were higher then a MIQM/NFS environment for all tests. A sample chart is shown below, for 2KB persistent messaging.
Controlled switch-over and fail-over times are faster with RDQM. The chart below shows much lower elasped times for RDQM, particularly for the fail-over test, where log replay to recover transaction state is against a local filesystem.
You can see the performance report for full full details of the tests measured, including these charts.