Ski> I am in the market for a new NAS system (30TB usable for end user Ski> use - not virtual machines) and our 2 finalists (EMC and NetApp) Ski> have taken very different approaches to solving the problem. I Ski> am wondering which solution you would be most comfortable with.
I'm curious about your needs here. Do you need 30Tb in one volume? Or are you going to carve it up into multiple volumes and then share them to the customers? And do you expect to have to grow/shrink volumes? I ask this because we're a Netapp shop and the 16Tb limit on aggregate size in OnTap 7.x, and the consequent limit on volume size, really really really kills us in terms of managing our disk usage. As project grow, I'd love to just be able to allocate them space, and as the project shrinks, I'd just shrink their dedicated volume in response. But since Volumes can't span aggregates, it's not a good way to work. This has made me look quiet closely at Isilon's (now EMC) product. It looks really good, but obviously it's not perfect. But having just a single volume image that you can grow on demand by just tossing new hardware into the mix and the underlying OS moves the data around for you... tempting. And the fact that you can mix and match their SATA and faster storage arrays to make a nice layered setup. But!! I haven't a clue how well their NDMP support for backups is. Ski> $Vendor1(more expensive): Use 256GB flash cache and 72 600GB 15K Ski> SAS disks (second option would have 96 450GB SAS disks if we felt Ski> we needed more IOPs). Minimal dependence on algorithms moving Ski> data between storage mediums. Ummm, just to make sure here, is the 30Tb you're asking for the RAW storage count, or the useable storage count? Netapp (and I assume EMC) are terrible at telling you what the useable storage will be for a system once you: A) add disks into a raid group (dual parity overhead) B) add raidgroups into an aggregate (5% snapshot by default) C) create a volume (20% overhead for snap reserve by default) So very quickly that 16Tb you started out with ends up around 12Tb and the boss is going WTF, where's all the disk space I paid for? So be careful to specify that they quote you 30Tb of NFS/CIFS free space once a standard setup is completed with their setup. Ski> $Vendor2 (much less expensive): Use 300GB flash cache, 8 - 100GB Ski> SSDs, 16 - 600GB 15K SAS, and 16 - 2TB 7.2K SATA disks. This Ski> depends then a lot on their technology for moving hot blocks to Ski> faster storage mediums. Ski> My environment does have a lot of data (e.g. student portfolio Ski> data) that is rarely touched so $Vendor2 may be a good fit. My Ski> concern is that a user will be working in a directory for 2 - 3 Ski> weeks, get used to a certain level of response and performance, Ski> then go to a directory that is on the slow disks and see a huge Ski> slowdown to the extent they think the system is broken. It's not quite clear to me what your usage scenario is. And possibly the killer isn't so much the storage, but the client OS issues and the need to train your users to split up their data a bit better so you don't have 20,000 entries in a directory! But from the sound of it, it's space, not performance that you need for a set amount of $$$. Then there's the hidden cost of managing it. I like how Netapp does stuff, and I can't talk to Isilon or EMC Celerra product. Ski> With our current NAS this is a big problem especially when the Ski> Mac clients open up a directory that has 20000 folders in it as Ski> Macs need all the meta-data on the 20000 folders before they will Ski> display anything in the GUI other than a beach ball. What is your current NAS? I suspect most NAS boxes are going to have trouble sending you the directory info for 20,000 items quickly, and that it's the CLIENT which is showing the issue here, not the NAs server. Segmenting your data out better would help the users and your performance quite a bit I'm sure. Ski> Has anyone had experiences with NAS systems that rely a lot on Ski> different storage mediums and migrating data between them to get Ski> performance? Appreciate your thoughts and ideas on this. We're currently using CommVault for backups and simple HSM of our Netapps, which makes things decent, but not quite as simple as I'd like. And it's not as quick as I'd like either, since the scanning process takes forever, but that's a function of the number of files we have on our filers. I was actually just thinking about storage management issues the other day because my users are, as always, filling up the disks and I've been hounding them to change their process to better reflect reality. I.e. they should have a simple, repeatable way to cleanup as they go along so that they don't end up with Terabytes of files they can't easily cleanup because they have no structure, etc. Which leads me to tools. I wish the storage vendors would add in some better reporting tools for helping manage disk space. One nice tool would be a simple list of the largest 1024 files in a volume. Easy to keep track, not alot of space in terms of memory or inodes, but absolutely vital when you need to quickly do a cleanup. Per-user du tree reports would also be nice, but a pain to keep upto date without impacting performance too much. Esp once you get into millions of files in a 20Tb volume. Hmm... I smell a LISA paper here somewhere, so I'll shutup for now. :] John _______________________________________________ Tech mailing list Tech@lists.lopsa.org https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
