| Performance of Remote Call [message #897607] |
Tue, 24 July 2012 13:13  |
Eclipse User |
|
|
|
Have there been any performance benchmarks established using ECF and the various providers... In the same application if I swap providers between ECF generic and r-OSGi, there is close to a 8X difference in performance.
using localhost and the r-OSGi, a simple remote call taking a couple of primative arguments takes approximately 200us. If I change the configuration to use the ECF generic provider, keeping the exact same source code - The remote call takes approximately 1500us to complete.
Any insights into the differences?
|
|
|
| Re: Performance of Remote Call [message #940315 is a reply to message #897607] |
Thu, 11 October 2012 10:23  |
| Eclipse User |
|
|
|
Walt Dixon wrote on Tue, 24 July 2012 13:13Have there been any performance benchmarks established using ECF and the various providers...
No, unfortunately not.
Quote:
In the same application if I swap providers between ECF generic and r-OSGi, there is close to a 8X difference in performance.
using localhost and the r-OSGi, a simple remote call taking a couple of primative arguments takes approximately 200us. If I change the configuration to use the ECF generic provider, keeping the exact same source code - The remote call takes approximately 1500us to complete.
Any insights into the differences?
There are a couple of possibilities. One is that this difference could be first-time only (i.e. upon first invocation only) and due to the more lazy loading and connecting of the generic provider. Do you see such a difference on every subsequent invocation?
If you would like to open a bug here
https://bugs.eclipse.org/bugs/enter_bug.cgi?product=ECF
we could investigate the differences and look to understand and/or eliminate them. There are ways to change/affect the generic provider's performance, if it's a problem.
|
|
|
Powered by
FUDForum. Page generated in 0.03836 seconds
Home