To summerise the thread so far and reply to some suggestions:
1) I should test minra off on the mirror target while doing a a backup. Good idea, I will try this. 2) minra can not be set on the target unless the mirror is broken or it is set on the source. This is just a fact of life at the moment. 3) the read ahead algorithm tries to speed things up for sequential reads. OK, thats fine, but backups are sequential reads, so you might think that turning on minra would make things worse for backups. But some peopel report the opposite. I have not seen a analysis from anyone that explains why minra has any effect on backups. Can somebody from netapp comment on this? Is the backup multithreaded? If not, I can imagine it being real slow for lots of small files no matter what you did. 4) People asked for details on the configuration. Its ONTAP 7.0.4, 3 shelves on the R200, single aggregate, 2 raid groups, 20 drives in each raid group, raid DPP, gig ethernet back to the backup machine, switched, not routed, NDMP in 3 way mode. Nice fast SUN running the backup software (netvault), backup goes to disk. Whew! For my database volumes I get 10-20 MBytes per sec sustained transfers. Not suprising, nice large files. For my other volumes, I get varying rates, but in all cases as the number of files goes up and the file size goes down by backup speed goes down. The mail volumes are obviously the worst case. The most interesting thing to me is item 3 above.
Regards, pdg
--
See mail headers for contact information.