Some picking of nits...
The market for NTAP is NOT EMC's customer base. It is the SUN and NT customer base. The market is expanding so rapidly, what the EMC customer base decides is almost irrelevant.
If the first sentence is true, then why do so many want to make this out to be an EMC v. NTAP showdown? I don't really believe that statement to be true, though. EMC, going down-market, and NTAP, going up-market, are bound to collide. The platforms that they attach to are irrelevant. EMC's customer base very definitely includes Solaris and NT buyers. The difference in their markets is in how the two devices are used, not who is buying them.
Suggest you study NTAP's architecture a bit more thoroughly. Interface to the network to which app servers and clients are attached may be multiple e-net, ATM, FDDI conections. Usually the app servers share one or more "dedicated" segments, isolated from the client LAN. More importantly, connection to storage devices is through FC or SCSI. (We are waiting for a secret joint development from BRCD and NTAP any day now.)
I've done enough work with NTAP to feel comfortable discussing their architecture. The pipe to them is plumbing, not architecture, though. IMHO, we can throw out the FDDI and ATM interfaces, because the whole point of the "argument" is the projected storewidth paradigm that is based on the standardization of Ethernet and it's coming 10GB flavor. Now, if we have a dedicated network segment (of whatever flavor you choose), are we not then building a seperate network for storage, something that the NAS people rail against when talking about SANs? I don't have anything against that approach, because it makes sense in high-volume environments.
Its been done and in production at customer insistence and driven by NTAP's cooperation. So far, it has been EMC that fights these implementations.
Now that is a smart customer getting the best of both worlds. And I can certainly see EMC fighting it. You sound like you or someone you know has "experienced the EMC effect." Their sales guys don't budge an inch, and their engineers budge even less. That's smart business, but it doesn't help out the visionary customer very much. It makes total sense to me that NTAP would want to cooperate in this kind of configuration, also.
There are many ways to skin the multi-TB cat. A filer farm of TB filers with or without clustering provides great scale. The idea of multiple TB being controlled by a single EMC is the storage management equivalent of mainframe computing. It isn't necessary.
I concur with the implementation of a filer farm as a cat-skinning methodology. That makes sense and does a job. I disagree with the inference that multiple TBs being controlled by a single EMC is "not necessary" just because it reflects a mainframe point of view. It's when you start scaling past the point of storage controlled by a single ANYTHING that I am talking about.
I've slowly (as usual) come to the realization that we are talking about two different market segments. My view of the market is that NAS filers are an excellent choice for low-volume storage, and EMC (and IBM, Hitachi, et al) are the right choice for high-volume storage. The split probably occurs today around the 10TB (that's a SWAG, based on end-user conversations) delineation between low- and high-volume data. Again, IMHO, it's the management of this data that will define the market more than architecture.
BTW, a whole bunch of ex-EMC engineers are at NTAP now. The experience to build multi-TB filers is there and will happen.
No doubt. They probably also have former Seagate, STK, IBM, Auspex, fill_in_the_blank engineers. The storage planet is a pretty small one in the big Computer universe. I would expect that EMC could point to a few former NTAP people, too. Each one of those individuals, no matter which company they're at now, could give you an iron-clad reason for moving from one to the other. In the same vein, if you personally are selling storage today or sometime in the last 10 years, you and I have probably bent elbows a few bar stools from each other, or passed each other in the halls of VeryLargeConglomerateCo.
I still think that the best thing for both companies is to work together at SOME level to grow the storage market. The synergies of the two technologies and architectures are simply too great to ignore. Someone, somewhere with more money and smarts than me will see this, and make a killing. Maybe they'll give me a job. <g>
buck |
