In watching ntop run on my sparc with a ~40Mbps (7500pkt/sec) or so stream
I noticed a large number of system calls and pretty sluggish performance.
Memory isn't an issue but a truss showed it calling time and getpid
thousands of times. Near as I can tell the tryLockMutex in the per packet
mutex is calling getpid forevery packet. As I understand it ntop is
threaded with the only fork being for static web serving so why do we
need to check the pid so much? Also the time call seems to be just for
accounting purposes so why not put it in the ifdef debug wrappers?
Just trying to squeeze some performance out of my old sparc hardware.
-Chris
BTW If anyone cares it a dual (soon to be quad) 400MHz ultrasparc with
2gig and a ZyNX quad 100meg ether.
--
[EMAIL PROTECTED] Chris Turbeville NTT/VERIO
Send mail with subject "send PGP Key" for PGP 6.5.2 Public key
_______________________________________________
Ntop-dev mailing list
[EMAIL PROTECTED]
http://listgateway.unipi.it/mailman/listinfo/ntop-dev
