On 01.01.2014 19:55, Paul J Stevens wrote: > Aleksander, > > imapproxy appears quite pathetic! > > I just did a simple benchmarker which does a login/logout cycle. > > hitting dbmail directly is an order of magnitude faster than hitting > imapproxy. That seems quite surprising since imapproxy doesn't have to > do any authentication, just hash and compare to the connection cache. > But still, the numbers speak clearly: > > testing login/logout > > COUNT | 10 | 100 | 1000 > ---------+--------+--------+--------- > DBMail: | 0.0532 | 0.6054 | 6.9042 > IMAPProxy| 0.3995 | 4.0095 | 40.0218 > > > I'm adding the benchmark script I wrote to the master branch so you can > test for yourself. > > http://git.dbmail.eu/paul/dbmail/tree/test-scripts/imapbench.py > > Note that for dbmail-2 using imapproxy made sense in order to reduce the > concurrency pressure. But for dbmail-3 that doesn't fly any more. I tried to update mysql but I'm not even close to get closer to the numbers quoted by Paul. Anybody can run imapbench.py on your system? I think 5 millisecondsis too fast for login/logout cycle. Maybe you are using something exotic, like database in RAM on localhost.
If it somehow help solve the problem, I can try to determine in which commit dbmail changed behavior towards imapproxy. _______________________________________________ DBmail mailing list DBmail@dbmail.org http://mailman.fastxs.nl/cgi-bin/mailman/listinfo/dbmail
