BCP IN and OUT Comparison between data in the SAN and MSA

  • I am wondering if anyone out there have had a comparison of extracting data in and out of a database wherein in one platform, the data resides on an MSA, and the other one is on the SAN. The data involved is almost exactly the same, there was just a little more data to extract out of the MSA, than with the SAN attached database server. The difference with the database server:

    * The one connected to the MSA is on a Windows 2003 server with SQL2005 in 32-bit.

    * The one connected to the SAN is on a Windows 2008 server with SQL2005 in 64-bit.

    Just to give an overview of the storage system:

    * The MSA has a data drive 8 disk subsystem configured as RAID 10. The log comprised of 4 disks configured as RAID 10 as well.

    * The SAN (which I don't really know much of the details other than it has 15 disk in the aggregate and is used for both Data and Log. It is using the NetApps filerview and doesn't really require having the separation of the data and the log. (I've read several debates on this about this, and it seems like a paradigm shift, i haven't made the shift yet, but i haven't seen any topic yet to invalidate it, for NetApp at least). That is another topic in itself.

    Okay, as for the BCP IN and OUT operations, it is showing that the BCP OUT for the MSA outperforms the operation for the SAN. I was expecting the other way around. What could be the explanation to this? Am I to conclude that the one with MSA is by far better than the one connected to the fiber channel?

    Please see Excel attachments as to the comparison of this.

  • Comparing I/O rates between two complex systems is tricky. What are the bus speeds for the two servers? How much cache on the two drive systems? Spindle speeds on both? I/O activity concurrent with the two operations? And a million other questions.

    Property of The Thread

    "Nobody knows the age of the human race, but everyone agrees it's old enough to know better." - Anon

  • Ayayay!

    I guess I am in the wrong belief that fiber-channel attached are always faster than other disk subsystems (except SSD).

    I guess, i was just really surprised. I really thought that SAN (especially NetApp) is always, hands-down faster, no questions asked, compared to non-SAN subsystems.

    Anyways, the only known factor is :

    For the MSA, the Data drive are 8 15k-rpm disk configured as raid 10 attached to a Proliant DL580G5 system (using Smart Array P800 Controller)

    For the SAN, the disk is a carving from an aggregate with 15-disk. Fiber-attached. 🙁

    Very few information to do the comparison with. But it was still surprising that the SAN was outperformed by our older, disk subsystem. 🙁

  • There is no reason that a properly configured DAS (direct attached storage) system like the MSAs can't perform as well or better than a SAN. In my experience with MSAs the controller and/or I/O bus is normally the bottleneck. The newer P812 controllers are significantly faster than the P800 controllers. (They have optimized sequential RAID5 and RAID6 writes so they are about as fast as sequential reads, I suspect blowing away RAID10 speeds, but I haven't gotten a chance to test that yet.)

    It seems you are more likely to get better performance with DAS as you have more control over it and you know that nobody else is using the same spindles as you at the same time. A properly configured SAN can do just as well, but it seems that often they have spent so much money on the SAN itself and believe all of the SAN marketing that they don't buy enough spindles and separate things out like they should.

    In reality, assuming everything else is equal, the 64-bit server should have a fairly big advantage from my experience.

  • Fibre chanel will just set a maximum transfer rate, not a minimum.

    In this case you just proved that your older MSA with 8 disks is much better at reading than your 15 shared disk SAN.

    Finding out why your SAN is much slower is not going to be easy, if you really want to find out you could just run a set of SQLIO tests on both MSA and SAN.

Viewing 5 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic. Login to reply