On Mar 4, 2006, at 7:47 PM, peter royal wrote:
On Mar 4, 2006, at 12:36 PM, [EMAIL PROTECTED] wrote:One thing we did was replace the MINA ThreadPoolFilter with our own threadpool filter that uses the Java 5 ThreadPoolExecutor.Aside from how you had your ThreadPoolFilter push write events into the event pool, how did you find performance? I just did a sample implementation that uses the TPE, and the performance for the regression test shows it to be slower than the Leader-Follower pool.. Any thoughts on how to improve TPE performance would be appreciated.. https://issues.apache.org/jira/browse/DIRMINA-184
I brought performance to be on-par with the Leader/Follower pool by using a BlockingQueue to hold per-session events, and allowing for a delay in reading events from the queue.. that is, by waiting a few ms for an additional event to arrive, performance stays up, since you save yourself re-queueing the session.
-pete -- [EMAIL PROTECTED] - http://fotap.org/~osi
smime.p7s
Description: S/MIME cryptographic signature
