Jeff probably hit the point - NetApp do not have a solution for SAN and I think that NetApp should re-think their direction. With more and more vendors competing in the nas /san domain, these products will be better and provide more functionality and better performance. The current connectivity limitations will be eliminated. With fibre fabric you will not be limited to the current limited implementation. I think that NetApps products are great but I am disappointed that they don't plan for future competitive products.
-----Original Message----- From: Jeff Van Cleave [mailto:jeffvc@icopyright.com] Sent: Wednesday, March 22, 2000 6:13 PM To: toasters@mathworks.com Subject: RE: nas storage
NetApp is not designed to compete with SAN - and yes, you cannot say that 100mb ethernet will compete with fiber channel.
However, direct attach has limitations in the number of servers it can connect to. At last check, EMC can connect with 32 systems. Hitachi can connect to a large number with a channel box (I don't know the high end limit). But you are limited by the length of cable between the SAN and the server.
Basically we approached using NetApp as a replacement for all of our NFS servers, and to some extent our data servers. For direct attach, we will probably go with Hitachi.
I think in any environment, both SAN and NetApp can be used effectively. It is not really an argument of which one is better, but what is your functional requirement and how will those needs be met.
Jeff
__________________________________________________ Do You Yahoo!? Talk to your friends online with Yahoo! Messenger. http://im.yahoo.com
NetApp do not have a solution for SAN and I think that NetApp should re-think their direction.
NetApp isn't in the SAN business. Whether they should be is an interesting question to discuss, but it's certainly not the foregone conclusion you paint.
I think that NetApps products are great but I am disappointed that they don't plan for future competitive products.
Personally, I think NetApp remains on track to continue to offer competitive products for some time. As best I can see, the only threat that could really put a hurting on 'em would be if someone came out with a cheap PCI board that took standard SDRAM modules and provided fast-n-cheap battery-backed storage, then started working on the device drivers to get it used properly with a good filesystem. Something like Reiserfs has a lot of the performance features people like in WAFL. If it knew intimately about a solid RAID implementation and battery-backed cache, about the only other thing to add would be snapshots. Oh, and getting CIFS compatibility right isn't a whole lot of fun. I'd guess NetApp has a few years play in 'em yet; they really do have a large amount of tough engineering to try to reproduce. The Open Source swarms may be chasing them. Will they be caught? It'll be fun to see.
Some people are really convined that "SANs" are a good idea. They string another complete network, generally alongside an existing IP network, typically using faster links, running proprietary protocols to provide remote simulation of the appearance of a locally-attached disk drive.
I've yet to see such a setting where they wouldn't have gotten better performance and functionality by running IP over the new, higher-speed network, and use whatever protocols are the best fit for the data on a case-by-case basis. Once upon a time there was an "Network Disk" protocol for providing remote access to something designed to look like a raw drive; Suns used to use it for diskless booting, way back when. It wasn't terribly nice, though, and when remote filesystem access became available and useable for booting, nd disappeared without a trace and nobody missed it.
I'm glad NetApp isn't trying to shove into the SAN fray; leave that turf to the aging, overpriced dinosaurs, desperately searching for new markets now that fewer and fewer customers will pay their premium prices for drive farms to attach to mainframes.
-Bennett
We have a Netapp 760 - we are using it with NT, Unix and even doing NFS mounts from OS390. We use it to glue Unix and NT apps.
We love it and will be getting another this year.
We also have a Brocade Fibre-channel switch and are doing fabric logins from SUN boxes and various STK disk. We are starting to see the real possibilites of fabric - with peripherals/hosts able to float anywhere on the fabric. For real - not glossy.
Fabric is the future - but everybody does SMB/NFS NOW. It will a few years before SAN has anything like the out-of-the box usability that NAS has right now.
* James A. Klun Sterling Commerce 4600 Lakehurst Ct. Dublin, Oh 43016 * * jklun@stercomm.com 614-793-7183 voice 614-793-7092 fax * * ---- "very like a whale" ---- * * ---- Hamlet, Act 2, Scene 2 ---- *
Fabric is the future - but everybody does SMB/NFS NOW. It will a few years before SAN has anything like the out-of-the box usability that NAS has right now.
Or before SAN has anything as flexible as NetApp. The 3 SAN presentations I have seen all had one big box of drives that got allocated out at the physical drive granularity and are a pain to enlarge space for a host. You can configure a RAID 5 set within a SAN and allocate it out, but you can't grow that dynamically. Have to unload the data, reconfigure a new set of drives, reallocate them to the host(s), then reload the data. Or burn another parity drive setting up a whole new RAID 5 set as well as configuring the host to recognize another drive device/mount point.
With a NetApp, you can allocate space at the MB granularity with quota trees, let everyone mount the whole filer, or anything in between. To add space, add a drive or a shelf of drives. Yes, you may end up with another parity drive depending on how many drives you're adding (I don't recommend 48 drives in one Raid group :) but that's optional.
Nope, for flexibility (and ease of administration), you just can't beat 'em! Just 2 more cents...