MQ

 View Only

MQ for z/OS - performance report for MFT

By Anthony Sharkey posted Mon February 22, 2021 08:01 AM

  

Almost too many years ago to remember, I created the "WebSphere MQ File Transfer Edition Performance report", also known as FP11.

The FP11 report provided information on the cost of managed file transfer when the source and/or destination transfer agents were hosted on z/OS, as well as considerations for tuning both the agents and MQ to achieve the best performance possible.


Over the intervening years, FTE has become MFT and many new features have been added, but the core function remains the same, i.e. to transport files from one system to another.

An updated performance report, "MFT for z/OS v920" using the latest long-term release code (v9.2.0) is now available.



The feedback we have had suggests many users run with the default agent.properties file options, so in this release we have concentrated on the basic transfers including:
  • Outbound from z/OS (dataset and USS files) to xLinux
  • Inbound to z/OS (dataset abd USS files)
  • LPAR to LPAR (whether z/OS datasets or USS files)

The performance report indicates a peak transfer rate of up to 250MB/second over a single MQ channel.  With a more modern disk subsystem, we would expect this peak rate to increase.

Similarly, the xLinux partner is a relatively old machine and is not necessarily indicative of the performance on the later more powerful systems, but it does offer some indication of the MFT performance even on older technology.


We also look at the benefits that the z15's on-chip compression has on MFT workloads, plus what improvements may be seen when transferring files over high latency networks with the fasp.io gateway.


It is intended that the document will see a much regular update cadence than FP11 when we see performance improvements to these key transfer types, whether from hardware or software updates.


This report, along with many other performance documents can be found on the mqperf github repository.






#IBMMQ
#MQ
#performance
0 comments
135 views

Permalink