This is a resend from about 2 weeks ago for which there was no reply. Any help is appreciated....
................................................................... I am incorporating ClamAv virus scanning into software that transfers files between two domains of different security levels. I have built the latest software from source, version 0.97.4, and have it running successfully for a slightly different application. I am using clamdscan to process whole directories to enable use of the multithreading capability. The host system is an older SGI Origin 3000 with about 200 processors running Trusted IRIX with a shared file system. What are some recommended configuration settings for clamd.conf for optimal use of multithreading? Realistically, how many threads can I expect to be able to make use of? In preliminary testing it takes about 30 seconds to scan 1500 typical transfer files. If I split these into two directories of about 750 files each the wall time for the individual scans are about 13 and 17 seconds for each (the difference is the volume of bytes in the two directories). However, when I run the two scans concurrently the wall time for each is about 37 seconds. This would seem to suggest that the clamd daemon is essentially timesharing between the two requests. True? I then created two separate clamd.conf files using separate Unix sockets and log files so I can start two separate clamd daemons. Running two concurrent clamdscan (each pointing to a different clamd.conf file) still does not produce improved (nor acceptable) walltime scan times. I would guess that I'm hitting some common resource bottleneck. Any ideas? --Dave _______________________________________________ Help us build a comprehensive ClamAV guide: visit http://wiki.clamav.net http://www.clamav.net/support/ml
