November 21, 2002 at 8:43 am
Is anyone know how to move existing sql server 2000 to SAN?
thank you advance for help
November 21, 2002 at 11:48 pm
Detach the Database and attach it. Or you
could also backup and restore the database.
Which SAN box have you guys bought. We are
in the process of implementing an EMC SAN.
SAN just provides you the storage media. Do
you plan to use the existing server to attach
to SAN or use a different server ?
Email me if you need more info.
Thanks,
rgn
November 21, 2002 at 11:59 pm
My personal experience with SAN's is not very good, especially in the speed department. SQL was very unstable on the SAN, but I must add that it is clusters we normally use on SAN's. But then again I have only worked on Compaq SAN's.
November 22, 2002 at 10:39 am
raino -- I would be interested in hearing about the performance issues on the SAN solution. We are in the process of migrating to an EMC SAN and I would like to know of others peoples issues so that are not surprises to me. Thanks.
David
David
@SQLTentmaker“He is no fool who gives what he cannot keep to gain that which he cannot lose” - Jim Elliot
November 22, 2002 at 11:58 am
Well we have mostly IBM Shark SANs, we do have a Hitachi and a small EMC.
So far, other than some setup issue's I have found them execellent for reliablity.
We have an engineering group that builds the servers for us so I am not heavily involved in that. The disks must be setup as Basic, and for some reason we have to use the slower IBM HBAs, versus the Compaqs that match the server.
One note is that with the fiber cards it is only marginally faster than SCSI, BUT you can have multiple HBAs and it then begins to outrun SCSI quickly. Additionally those multiple cards provide redundancy and load balancing.
KlK, MCSE
KlK
November 22, 2002 at 1:04 pm
We are using an EMC SAN for three instances on clustered servers, including SAP. Our performance is pretty good and have even migrated databases and servers to it, just to increase performance.
So far I am happy with EMC.
"Keep Your Stick On the Ice" ..Red Green
November 22, 2002 at 1:06 pm
I forgot to add that I have had problems that were solved by increasing the number of threads I have to the file groups. Make sure you have a good physical architecture with your databases and I bet the EMC will do just fine....
"Keep Your Stick On the Ice" ..Red Green
November 25, 2002 at 6:57 am
We're running Compaq 8500s, dual-attach to an EMC Fiber Array (fabric), front-ending our IBM Shark and we have NO performance problems at all. (Win2k and SQL2k)
We 'carved out' 3 blocks of space and assigned them as DATA, LOGS, and BACKUP. It's our experience that breaking them out was unnecessary since ALL drives in the Shark act as one array, and we would really only gain performance if these 3 resources were on different physical spindles.
Hope this helps.
Eric
November 25, 2002 at 9:04 am
We have a Hitachi SAN with 10TB of disk. The SAN performance is signifigantly greater than any of the local disk arrays. It has also simplified our lives as we no longer need to configure separate arrays for data and log, it just didn't make much difference.
We have use single and dual HBA cards, but have never run the dual cards in a load balancing mode. The clustered servers use single cards and non-clustered servers use dual cards in failover configurations.
Our only problems to date have been:
a) Do not boot up many servers at one time or random SAN drives will be missing.
b) Do not do configuration changes onthe SAN during the day as it can cause SAN drives to dissapear.
c) Just because you can shine a light through it does not mean a fiber cable is good.
November 26, 2002 at 11:02 am
Thank you! Every One!
November 27, 2002 at 3:25 pm
we're running a 1Tb emc/clarion san with brocade switches. The performance is excellent. Someone said earlier that you couldn't run the disks as dynamic but all of mine are - no problems. my clearpath mainframe is attached as well. Carving multiple blocks out a one array is helpful if your os is capable of firing io's down multiple paths.
November 28, 2002 at 2:37 am
I have several databases on EMC SANs. My only emphasis is to be sure to physically design your database files as you would on any other RAID or set of disks. Our original BIN design was not designed with physical IO in mind, and indeed, we ran into IO bottlenecks. This was mostly due to standard SQL Server backups going to the same disks where data and log files also resided (data, log, and backup meta and hyper volumes were configured to the same set of physical disks). Bad design, but my point is that IO architecture should not be overlooked. SANs and large IO cache is not a magical end to IO issues.
Take a look at snapshot backups to possibly eliminate standard backups, which would also reduce, if not eliminate, typical IO associated with database backups. BCVs are another big advantage of EMC SANs. BCVs can offer an easy way to maintain offline databases for reporting, disaster recovery, etc.
November 28, 2002 at 9:18 am
Pardon for the late post, but this is an interesting thread and I'd like to see it stay alive a little longer.
I've worked with a couple of SANs.
EMC Clarion was quick to come on line and the few issues were solved quickly by suport staff. The performance was much better than the previous RAID storage simply because of the faster transfer rates and the huge cache at each controller. We did observe some throughput advantage when the log and data were directed through spearate controllers in the EMC box when running under stress, but nothing huge.
Currrently, I am working in a mostly SCSI environment and we have been attempting to get databases migrated to an HP SAN. The hardware has been in the building for several months. We are continually having problems that even the HP techs can't resolve. I fear for the day they actually ask me to fire up a production database on this hardware.
Another interesting problem I am facing is tuning the stripe size of the SAN and SQL Server. I saw a test in Niel Pike's FAQ on a SAN and SQL 6.5 to determine an optimal stripe size. I've been hoping to dupplicate the tests in my environment and SQL 2000, but so far we can't keep the SAN up long enough to actually experiment with changing the stripe size. Anybody have any experiences to share?
Bill Wunder
November 28, 2002 at 5:01 pm
We have three clustered servers with their own external storage arrays as well as a clustered SQL Server install where the drives are part of our SAN. The SAN-based one hasn't been promoted into production because of a delay in the project itself. So under little or no stress there weren't any issues, but I can't speak to what it would be like with a load on the system.
We're all Compaq and we've had one major issue with one of the shared storage clusters, but it was only on one of the nodes. Replacing the server itself and doing the rebuild of the cluster resolved the issue. The others haven't had any issues as far as SQL is concerned. A vendor had come in to setup the initial cluster and we did have issues with their setup on two of the other clusters.
We also did have an issue with a clustered server instance providing file services from the SAN, but it may have been related to some quota software that had been installed (Microsoft's engineers were clear that most quota software aren't supported). However, we took a shotgun approach due to the criticality of the cluster so it's hard to say what solved the issue.
K. Brian Kelley
http://www.truthsolutions.com/
Author: Start to Finish Guide to SQL Server Performance Monitoring
http://www.netimpress.com/shop/product.asp?ProductID=NI-SQL1
K. Brian Kelley
@kbriankelley
September 16, 2003 at 3:20 am
Hi all
A couple of things are missing in this posts that I would like comments on...
a) IBM Shark - Supports only RAID-5 or can it do RAID-10? are people running RAID-5 array's primarily within the SAN and has anyone experience re performance?
Just that RAID5 writes are v.slow and thought this would be a major issues in the SAN?
b) SAN Cache - one ive seen has 8Gb, the split is 384Mb write and the rest for read, the SAN is this case is RAID-5, therefore, it doesnt seem correct to me. I would opt for a 50/50 or perhaps a 60read/40write (minimum). Thoughts on this??
Cheers
Ck
Chris Kempster
Author of "SQL Server 2k for the Oracle DBA"
Chris Kempster
www.chriskempster.com
Author of "SQL Server Backup, Recovery & Troubleshooting"
Author of "SQL Server 2k for the Oracle DBA"
Viewing 15 posts - 1 through 15 (of 16 total)
You must be logged in to reply to this topic. Login to reply