Richard Elling wrote: > Torrey McMahon wrote: >> Spencer Shepler wrote: >> >>> On Jul 10, 2008, at 7:05 AM, Ross wrote: >>> >>> >>>> Oh god, I hope not. A patent on fitting a card in a PCI-E slot, >>>> or using nvram with RAID (which raid controllers have been doing >>>> for years) would just be rediculous. This is nothing more than >>>> cache, and even with the American patent system I'd have though it >>>> hard to get that past the obviousness test. >>>> >>> How quickly they forget. >>> >>> Take a look at the Prestoserve User's Guide for a refresher... >>> >>> http://docs.sun.com/app/docs/doc/801-4896-11 >>> >> >> Or Fast Write Cache >> >> http://docs.sun.com/app/docs/coll/fast-write-cache2.0 >> > > Yeah, the J-shaped scar just below my right shoulder blade... > > For the benefit of the alias, these sorts of products have a very limited > market because they store state inside the server and use batteries. > RAS guys hate batteries, especially those which are sitting on > non-hot-pluggable I/O cards. While there are some specific cards > which do allow hardware assisted remote replication (a previous > Sun technology called "reflective memory" as used by VAXclusters) > most of the issues are with serviceability and not availability. It > is really bad juju to leave state in the wrong place during a service > event. > > Where I think the jury is deadlocked is whether these are actually > faster than RAID cards like > http://www.sun.com/storagetek/storage_networking/hba/raid/ > > But from a performability perspective, the question is whether or > not such cards perform significantly better than SSDs? Thoughts?
In the past we saw some good performance gains with on server caching of the i/o even if it went to a large array with lots of cache down the road. Mr. Sneed had some numbers from back in the day but for some workloads there was quite the impact by ack'ing writes and coalescing them on the host NVRAM before sending them to the array. You cut the i/o time down for the app *and* the i/o you do send is a lot less bursty. (Of course if you filled up the memory on the card you'd hit other issues.....) _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
