Memory use

Ben Wilhelm zorba at pavlovian.net
Thu May 26 19:24:21 UTC 2005


0.1.0.5 and 0.1.0.8 both show this behavior. I haven't been doing 
anything heavy since post-0.1, so I don't know if it's gotten "better". 
This is acting as a client, although I've noticed it devours quite a lot 
of CPU when acting as a server also. Mostly I'm concerned about the 
client right now. :P

-Ben

Jonathan D. Proulx wrote:
> On Thu, May 26, 2005 at 11:45:42AM -0700, Ben Wilhelm wrote:
> :
> :It also uses a metric ton of CPU power when it's under heavy load. I've 
> :had it saturate the entire CPU, and I think it stops transmitting data 
> :in those situations too. It's spending a lot of that in a giant 
> :linear-time search through a bunch of strings, and when I get some free 
> :time I'm going to try to speed that up to O(log n) or a hash table or 
> :something, and send the patch in if it seems stable and good.
> 
> are you running 0.0.9.x or 0.1.0.x 
> 
> the 0.1 stuff has been using a lot less CPU, though it could still be
> a lot for some.  My mixed use dual xeon workstation (spoon) runs at
> about 30-40% CPU for TOR while a dedicated Sun E250 (kropotkin, dual
> UltraSaprc II) running Linux sweats at 50-60% CPU (for TOR) and 480M
> memory utilization (system wide).
> 
> These are heavy specs, but the 0.0.9 stuff would happily suck down all
> the CPU on my worksation and start to choke out interactive
> processes.  If you want to cut down on CPU beyond this you can tweak
> "MaxAdvertisedBandwidth" in torrc
> 
> from the torrc man page:
> 
> MaxAdvertisedBandwidth N bytes|KB|MB|GB|TB
>               If set, we will not advertise more than this amount of
>               bandwidth for our BandwidthRate.  Server operators who
>               want to reduce the number of clients who ask to build
>               circuits through them (since this is proportional to
>               advertised bandwidth rate) can thus reduce the CPU
>               demands on their server without impacting network
>               performance.
> 
> Don't know how if this influences memory
> 
> -Jon



More information about the tor-talk mailing list