JMS Perf Cases

Added by willpugh, last edited by willpugh on Sep 28, 2004  (view change)

Test Cases

This is a brain dump of the JMS perf scenarios I think should be tested. This is still really rough. I need to fill in these cases better, and incorporate a little more of the info in Things To Consider. I'm hoping to get a bit more JMS expertise in here to help out flush out these cases.

Measurements

For all of these, the main measurements we will be making in V1 of the framework are the throughput of the system and the average message processing time. I expect people will want to set-up performance monitors on the JMS machines to get information about disk activity, working set, etc. Unfortunately, we will need to push these off automating these measurements until a later version of the tool.

Cases
  1. Point to Point message, non-durable, non-clustered
  2. Point to Point message, non-durable, clustered
  3. Point to Point message, durable, non-clustered
  4. Point to Point message, durable, clustered
  5. Pub-Sub with one publisher + a couple of subscribers, non-durable
  6. Pub-Sub with one publisher + a couple of subscribers, durable
  7. Pub-Sub with a couple of publisher + a couple of subscribers, non-durable
  8. Pub-Sub with a couple of publisher + a couple of subscribers, durable
  9. Pub-Sub with a many publishers + many subscribers, non-durable
  10. Pub-Sub with a many publishers + many subscribers, durable
Issues
  1. So when I was looking at this, I was thinking that these test would not be going agianst embedded JSM server, because I didn't think that occured to much in reality. Does it?
  2. I think we need cases in here that take advantage of having a broker

Labels

 
(None)