Last night’s backup ran with 8 concurrent streams. The time for Agent to Tape was 3 hours exactly. An identical traditional backup two weeks ago took 13.5 hours. One interesting thing: The same two machines had TCP communication timeouts again. I dug back further in the logs and found they’d been doing this for some time when there was only 1 stream for agent-tape backups. Their problem is independant fo multiple streams. I’m switching back to 15 streams and will wait to see how it works out this weekend with a full backup.