Thank you Ales, what you write makes sense to me. I did some testing with some surprises:
1. when MAs and SOS storage are on the same client with the tape device, increasing number of MA increases throughput up to a point. I can get ~110 MB/s with
2. when MA and SOS is on client connected with 1GBit network to client with tape, I can get best performance with 3 MA (75MB/s), increasing number of MA actually decreases throughput. Is seems that fight for the bottleneck resource (network) has high overhead.
3. in both cases above I can see the SOS process reads 2-3x time more data from disks that are actually passed with MA to to target. Is this expected? It seems like SOS does some kind of read-ahead but it misses the target.
4. with 1 and 2 I decided to decrease to decrease concurrent streams on tape device from 10 to 5 and I expected to see tape write around 110MB/s from local SOS + 75MB/s from remote SOS = 180 MB/s. However I was surprised to see write speed of 40 MB/s, in other words same speed as for 10 concurrent MAs. I have no idea what's wrong and the performance is bad.
Can you help me? Do you suggest me to raise case with HP support?
Kind regards, Jan