I’m copying couple of hundreds G of data around and it is slow, really slow.
I know I made something wrong – I was thinking concurrency is the bottleneck as single copy would not be able to maximize CPU and network, however, after I launch 30 copy at the same time, I found … the bottleneck is actually coming from disk.
SATA sucks, it is just an enhanced IDE as someone told me before. 30 copy running at the same time caused all processes are waiting for IO, and thinking of throughput is almost nothing, it seems to me it is waiting for disk seeking.
I still need to do some tests and then gather statistic information to see how much the SATA can do – I was running 20 copy on a SCSI (well, RAID’ed) and it was doing great.