Due to a prospect/POC activity a customer is thinking about performance issues in order to decide to adopt (or not) ServiceMix 3.1.x in production.
Performance constraints in production:
1) 12 Million HTTP hit / Day, on simple processes, composed by:
- 1 Jetty BC (MEP: IN/OUT, NO SOAP)
- 1 Lightweight component, which generates XML output using
java StringBuffer Object.
- average response msg size: 10KByte
70% of daily load concentrated in 6 hours, corresponding to
390 HTTP Hit/sec.
2) Expected peak of 1000 HTTP Hit/sec, generated by 1000 concurrent
3) Architecture: horizontal cluster (only for load balancing) composed
by 4 independent nodes (ie: not in ServiceMix cluster mode).
4) Each cluster's node load capability must remain under 80%,
running under stress.
So, for each node (server) corresponding values are:
- 100 HTTP hit/sec
- 250 Hit/sec peak
- 250 concurrent client (peak)
As I said, because of high hit rate and corresponding message size, the
customer is warried about performance. He gave us a benchmark published by WSO2 (http://wso2.org/library/2259 : SCENARIO 1), which put in evidence ServiceMix 3.1 performance capability drop when number of clients grows.
Basing on info from such benchmark (tailored on dual core 2,0 GHz machine) the customer is afraid to be in a "borderline" position, having this hardware: 4 VMVare omogeneous machine each one with
IBM System X3550 4 Xeon 5160 3GHz: (SPEC index per core = 30,75).
It would be great to be able to provide assuring infos (better referencing his Hardware), in order to persuade the customer to adopt ServiceMix solution for this production rollout.
Someone have this kind of information, and can provide us ?
Thanks in advance