JXInsight/OpenCore 6.4.EA.2 Released – Adaptive “Response (Time)” Control
In the second early access build of JXInsight/OpenCore 6.4 “ACE” we have added a new metering extension,
rvalve, that self regulates the concurrency of probes associated with a valve for the purpose of response time optimization (and indirectly throughput) using various flavors of local search (optimization) including hill climbing, reactive search optimization (RSO) and tabu search. This
rvalve extension is similar to the
tvalve extension and in many cases will have a similar controlled influence on performance of an application. The primary difference is being the measure used to compare the performance under different concurrent workload conditions. The
rvalve uses the mean response time whereas the
tvalve is largely driven by the throughout. Whilst both extensions could in theory be used in general one or the other will be enabled within the metering runtime.
Note: Research has shown that the throughput measure can be more accurate when the service time has high variability and the load is intermediate though the last one has to some degree being addressed in the
rvalve controller implementation.
Below are two throughput metric charts showing the concurrency hill climbing performed by both the
rvalve metering extensions in executing the same highly saturated test case. The drop in the middle is the completion of the first run followed by the creation and starting of new benchmark threads.
Further information on the large number of control set points offered by the
rvalve probes provider metering extension will be published shortly on our developer site.