Before we put our T2000's into service as Netbackup master servers, I benchmarked their network throughput using ttcp and it was amazing. We pushed 5 simultaneous ttcp streams into the T2000 for a cumulative throughput of 495 MB/sec (that's megabytes). All of this was without it breaking a sweat on the CPU--quite low utilization. Take careful note that we were using the 6/06 version of Solaris 10; it takes advantage of the GLDv3 rewrite of the Solaris 10 networking stack, including network link aggregation.
We put it into service as a Netbackup master server and we aggregated 6 incoming GigE links. Like the post above, I used 256K as the buffer size (recommended by the the LTO3 drive manufacturer, HP). I used empirical methods to find the miniumal number of buffers that still provided benefit. I doubled the number of buffers, tested, doubled, tested, etc. I found that 128 buffer provide a slight increase in speed vs. 64 buffers. But beyond 128 there was no gain. _______________________________________________ Veritas-bu maillist - Veritas-bu@mailman.eng.auburn.edu http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
