-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Toad wrote:
> On Mon, Jul 28, 2003 at 05:25:01PM +0300, Jukka Holappa wrote:
>>My node is serving > 20000 requests per hour and is not even overloaded
>>yet! With stable build about 6000-8000 requests hits the CPU limit.
>>
>>It also seems to find a lot of working nodes and use them:
>>Total Trials  230502
>>Total Successes       182887
>>
>>Previously the successes-value was a lot less than total.
>
>
> I wonder what part of this is due to nonblocking QueryRejected sending,
> and what part is due to open connections only routing...

I assume it's mostly open connections only routing that makes the
difference. I didn't have that many threads used to QueryRejects.

My current thread usage is the following:

Total pooled threads      522
Available pooled threads  101
Pooled threads in use     421

Configured maximum is 350 threads, I'm almost constantly queryrejecting
now.  Local mean traffic is ~35000, but my CPU is still not used up.

Checkpoint: Fetching ARK freenet:SSK                     1
freenet.Message: ClientGet                               5
freenet.Message: DataNotFound                            3
freenet.Message: DataRequest                           142
freenet.Message: InsertRequest                           2
freenet.Message: QueryRejected                          26
freenet.Message: QueryRestarted                          8
freenet.OpenConnectionManager$ConnectionJob             23
freenet.interfaces.LocalNIOInterface$ConnectionShell     3
freenet.interfaces.PublicNIOInterface$ConnectionShell    6
freenet.node.states.data.DataSent                        3
freenet.node.states.data.DataStateInitiator            173
freenet.node.states.request.RequestInitiator            26

Normally most (usually near 200) threads are used to DataStateInitiator
and next one is always DataRequest (usually near 100). QueryRejected has
never been near these two.

I can only assume that these threads are waiting the transfer, since
they don't use CPU. I hope you get your phase 2 of NIO finished so we
can get rid of these bottlenecks too :)

I really can't increase much the maxthreads since it will hit the memory
limit (stack allocation), I'm already using smaller stack.. but I can't
make it much smaller before some of these threads hit the bottom of it :\

- - Jukka
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQE/JtCLYYWM2XTSwX0RAnw0AJkBkTjxsZtAoBlu2c2G/qDCxIMtAwCdF3Wm
cGuakzBoW29fpEOdU5ONSX0=
=Qdnp
-----END PGP SIGNATURE-----

_______________________________________________
devl mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl

Reply via email to