Greetings, all. I'm contemplating some upgrades to my TSM infrastructure, and trying to determine where I think the bottleneck will move.
I'm adding tape drives, and intend to do a huge load of reclamation Real Soon Now. So, my question is: Does tape-to-tape copy which the TSM server deems "local" actually touch the server bus at all? For example; if I have 32 3592 drives (shush; let me dream), sufficient SAN fabric to keep them all fed, and sufficient reclamation work to keep them occupied, all totally devoted to reclamation: Is my -server- doing I/O work, or just bookkeeping? The tape IO management stuff (or lack thereof) leaves me feeling adrift, when compared to the superb support for disk IO evaluation I've grown to love out of NMON. I feel like I understand my disk IO intimately, but really wonder what's going on with the tapes. I've tried to work this out by calculating the 'negative space', but when I'm doing lots of tape work, I'm also beating the snot out of my database, so I'm not sure where to blame the IOWAIT. Anyone got a clue to spare? - Allen S. Rout
