A recent article in Server/Workstation Expert, entitled "NAS Wars," might be of interest here:
http://swexpert.com/CB/SE.C11.MAR.01.pdf
I applaud Alan's efforts at applying a bit of math and a lot of common sense to such a sticky issue.
This isn't meant to start a flame war, though something I read as a kid on a package of firecrackers comes to mind: "Light fuse, and run away fast!" ;)
Personally, I would suggest reading the first three sentences of the conclusion first, and with that perspective, read the rest of the article. Read that way, the methods and assumptions and calculations and configurations make for good detective work, IMHO.
Until next time...
The Mathworks, Inc. 508-647-7000 x7792 3 Apple Hill Drive, Natick, MA 01760-2098 508-647-7001 FAX tmerrill@mathworks.com http://www.mathworks.com ---
----- Original Message ----- From: "Todd C. Merrill" tmerrill@mathworks.com To: toasters@mathworks.com; emcnas@mathworks.com Sent: Friday, March 23, 2001 3:01 PM Subject: NAS Wars
A recent article in Server/Workstation Expert, entitled "NAS Wars," might be of interest here:
http://swexpert.com/CB/SE.C11.MAR.01.pdf
I applaud Alan's efforts at applying a bit of math and a lot of common sense to such a sticky issue.
I just finished reading it. Not a bad article, really. But I must say it sounds like Alan had a pro-EMC bias. He admits to all the noted discrepancies between the EMC and the NTAP configurations, but he make no attempt to analyze them quantitatively; i.e. doubling the amount of EMC hardware to more realistically compare RAID-0 with RAID-4. He goes so far as to cast unfounded suspicions on Netapp; i.e. "Who knows (I am a suspicious sort), perhaps a scaling problem is why NetApps chose not to use more disks." He then launches into some dubious assumptions and calculations about disk ops to make the IP4700 look better than it performed on the Spec benchmarks, and ends with a completely unfounded "I would expect the IP4700 to have a significant advantage over the NetApps F840 due to what would seem a richer hardware architecture." Well, heck Alan, you would expect that, but you'd be wrong! He only begrudgingly admits that WAFL "may" have something to do with this!
He basically starts off by trying to question the validity of the Paine Webber report, finds difficulty in doing so, and winds up declaring that it's all the fault of lack of further information. (Alan probably thinks, "If I could just get more information, I'm sure I could show that the IP4700 beats the F840 in at least one juryrigged configuration!")
Bruce
PS - And he doesn't even mention cost. What if the F840 is half the price of the IP4700???
I agree with Bruce.
The author is a " a 20 year veteran", yet hasn't figured out that more hardware doesn't necessarily mean more performance.
How many times have I seen this!
Talk to anyone who has moved from Novell to NT, It takes more hardware to get the same speed.
Test Linux and NT on the same hardware, heck load up dos and run some test. You need a lot less hardware if the OS is tighter.
Move from a simple database to a relational one.
Netapp has Ontap - one real-time OS. IP4700 has Crosstor, Vxworks and the Clariion disk management.
Correct me if I'm wrong but one of it's processors is doing raid and I/O and the other is doing file system, protocols, and IP.
This isn't a bad thing, just that it's going to take a little more hardware to match up.
This article didn't talk at all about software, features or the implementation of NFS / CIFS, backup, card support, etc. just that more hardware should make a system faster.
--- Bruce Sterling Woodcock sirbruce@ix.netcom.com wrote:
----- Original Message ----- From: "Todd C. Merrill" tmerrill@mathworks.com To: toasters@mathworks.com; emcnas@mathworks.com Sent: Friday, March 23, 2001 3:01 PM Subject: NAS Wars
A recent article in Server/Workstation Expert,
entitled "NAS Wars,"
might be of interest here:
http://swexpert.com/CB/SE.C11.MAR.01.pdf
I applaud Alan's efforts at applying a bit of math
and a lot of common
sense to such a sticky issue.
I just finished reading it. Not a bad article, really. But I must say it sounds like Alan had a pro-EMC bias. He admits to all the noted discrepancies between the EMC and the NTAP configurations, but he make no attempt to analyze them quantitatively; i.e. doubling the amount of EMC hardware to more realistically compare RAID-0 with RAID-4. He goes so far as to cast unfounded suspicions on Netapp; i.e. "Who knows (I am a suspicious sort), perhaps a scaling problem is why NetApps chose not to use more disks." He then launches into some dubious assumptions and calculations about disk ops to make the IP4700 look better than it performed on the Spec benchmarks, and ends with a completely unfounded "I would expect the IP4700 to have a significant advantage over the NetApps F840 due to what would seem a richer hardware architecture." Well, heck Alan, you would expect that, but you'd be wrong! He only begrudgingly admits that WAFL "may" have something to do with this!
He basically starts off by trying to question the validity of the Paine Webber report, finds difficulty in doing so, and winds up declaring that it's all the fault of lack of further information. (Alan probably thinks, "If I could just get more information, I'm sure I could show that the IP4700 beats the F840 in at least one juryrigged configuration!")
Bruce
PS - And he doesn't even mention cost. What if the F840 is half the price of the IP4700???
__________________________________________________ Do You Yahoo!? Get email at your own domain with Yahoo! Mail. http://personal.mail.yahoo.com/
sly789@yahoo.com (sg) writes:
I agree with Bruce.
The author is a " a 20 year veteran", yet hasn't figured out that more hardware doesn't necessarily mean more performance.
Yes. The bits that caught my eye were
| Althought the Alpha is a 64-bit processor, NetApp's operating | system was just 32-bit, so they were not exploiting the power | of their design.
and
| [...] one could argue that moving down [sic] from the Alpha to | the Intel CPU was a significant technology and performance step | backwards, but [...] For instance, there are far more operating | systems based on Intel than Alpha, as well as applications | software for these operating systems.
One wonders whether the author has any idea how a NetApp box looks to the outside world. Does he really think that the architecture of the filer CPU is significant to its clients? Or that NetApp run MicroSloth applications on their (Intel-based) filers?
Now if he had made some serious points about NetApp's implementation being dependent on cost-effective fast uniprocessors, and what the future prospects for such beasts might imply (as compared to systems that can make better use of multiprocessors), rather than wittering on about irrelevancies like the above, maybe the article would have been worth reading.
Chris Thompson University of Cambridge Computing Service, Email: cet1@ucs.cam.ac.uk New Museums Site, Cambridge CB2 3QG, Phone: +44 1223 334715 United Kingdom.
----- Original Message ----- From: "Chris Thompson" cet1@cus.cam.ac.uk To: "sg" sly789@yahoo.com Cc: emcnas@mathworks.com; toasters@mathworks.com Sent: Sunday, March 25, 2001 2:46 PM Subject: Re: NAS Wars
Now if he had made some serious points about NetApp's implementation being dependent on cost-effective fast uniprocessors, and what the future prospects for such beasts might imply (as compared to systems that can make better use of multiprocessors), rather than wittering on about irrelevancies like the above, maybe the article would have been worth reading.
I agree. And that's a great point, Chris. I don't doubt the ability of Netapp to do multiprocessor, and for all I know they have already done the code. But their approach, so far, has always been to rely on the scaling of single processors, both in performance and in price, to keep turning their product line. Will there come a time when that will no longer be possible amd multiprocessor offerrings become necessary? Sure, but I thought that a few years ago, and it still hasn't happened.
The "dual everything" IP4700 design is nice, though... and with two processors you can still continue to function if the other one fails.
Bruce
I think the article brings out a good point. Why can't we get a simple one to one test of the two boxes withthe same amount of useable storage, and as similar a configuration overall as possible? That way we can really see the price and performance of each. Up to now all we have is hearsay and Marketing BS. I don't think we're comparing apples to apples yet. For example, I suggest we test five real world applications. 1) Test both in a home directory configuration 2) Test both running a common Oracle application 3) Test both running a Lotus Notes application 4) Test both running exchange 5) Test both simulating recovery form an Oracle outage and an Exchange outage and how long it takes to recover.
Let the best box win : ).
--- Bruce Sterling Woodcock sirbruce@ix.netcom.com wrote:
----- Original Message ----- From: "Chris Thompson" cet1@cus.cam.ac.uk To: "sg" sly789@yahoo.com Cc: emcnas@mathworks.com; toasters@mathworks.com Sent: Sunday, March 25, 2001 2:46 PM Subject: Re: NAS Wars
Now if he had made some serious points about
NetApp's implementation
being dependent on cost-effective fast
uniprocessors, and what the
future prospects for such beasts might imply (as
compared to systems
that can make better use of multiprocessors),
rather than wittering
on about irrelevancies like the above, maybe the
article would have
been worth reading.
I agree. And that's a great point, Chris. I don't doubt the ability of Netapp to do multiprocessor, and for all I know they have already done the code. But their approach, so far, has always been to rely on the scaling of single processors, both in performance and in price, to keep turning their product line. Will there come a time when that will no longer be possible amd multiprocessor offerrings become necessary? Sure, but I thought that a few years ago, and it still hasn't happened.
The "dual everything" IP4700 design is nice, though... and with two processors you can still continue to function if the other one fails.
Bruce
__________________________________________________ Do You Yahoo!? Get email at your own domain with Yahoo! Mail. http://personal.mail.yahoo.com/
-
Bruce Sterling Woodcock -
Chris Thompson -
grey jolly -
sg -
Todd C. Merrill