On 6/14/2011 9:41 AM, Adam Levin wrote: > > On Tue, 14 Jun 2011, Edward Ned Harvey wrote: >> Agreed fully about disbelieving the company line... >> And I would personally choose Oracle/ZFS in nearly every case over the >> Netapp. ;-) > > I'm very excited about ZFS. We've *just* began playing with it a little > bit. I'm not sure I trust it enough for production, but I'm expecting > very big things from it. > zfs is pretty good in production. We've been using it for 4 years. There are bugs in the OS dealing with zfs to be sure. The bug list is a bit scary if you look at it. It is reliabile but not as reliable as a Netapp, or a BlueArc for comparison purposes.
snapshots: essentially unlimited backups: very efficient because it knows what has changed integrity: This is the best feature. It will tell you when disks are lying to you about the data that they are returning. Disks "DO" lie, and it's not consistent. ZFS is an end-to-end filesystem that can correct errors in any subsystem on the fly and return the correct data to you. But: we still get an OS hang sometimes. Also, zfs is the most reliable part of the Oracle storage appliance. We bought a fishworks and were not able to put it into production for 2 years because of extremely serious bugs in replication, failover, clustering, and all that stuff. The newer ones are supposed to be better, but damned if I'll ever buy another one.. NEVER. On the other hand there are a bunch of companies based upon OpenIndiana and Illumos and more on the way that seem to have pretty good simple failover for zfs in a cluster (Nexenta, Berkeley Communications, others) > At one time I believe several other vendors (besides Oracle) were looking > at front-ending their SAN solutions with a BSD-ish box running ZFS for > file shares. I'm really curious how that'll pan out, especially for > performance. > > The only reason I kept my discussion to EMC and NetApp was because the > original message said that the process was down to those two vendors. > >> ZFS does storage tiering, but it's more along the lines of caching. The >> system chooses what will be in the various levels based on historical usage >> patterns. You don't get to choose or influence the balancing very much. I >> like more control knobs and gauges. > > While there's definitely something to be said for knobs and gauges, I have > to be honest: the older I get, the less work I want to do, and if the > array can be configured to do the work for me, yay! :) > one thing that zfs isn't good at is modifying the size of the pool after you create it. That's a pretty serious limitation these days. Sure, you can expand a pool, but things are unlikely to be quite balanced in the future. (different disk sizes, models, characteristics, etc.) You can manually do some stuff like opening, reading, and writing every file to rebalance, but it's a bit tedious. Other filesystems do this much better. >> Clearly my perspective has biased positives in favor of ZFS, but I haven't >> had any Netapp in a few years, and I don't know much else. Can anyone shed >> more positive light on either Netapp or anything else? > > Well, all I can say is that I think ZFS, as it matures, is going to become > a very serious game-changer in enterprise storage and is a very strong > competitor for NetApp, especially considering the costs. I think your > summary is good -- ZFS does have a lot of benefits that they bring to the > table. If the performance can stay good, they're going to be a very > strong player in the market, I think. > > -Adam > > _______________________________________________ > Tech mailing list > [email protected] > https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech > This list provided by the League of Professional System Administrators > http://lopsa.org/ _______________________________________________ Tech mailing list [email protected] https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
