XPS or IDS customers using SANs? - Results

IBM via IIUG recently requested information on the use of SANs by IDS and XPS customer base. The survey was simple and consisted of 6 questions: - Who is the customer? - On which platform is the server on? - What storage vendor is being used? - How much data is being storing? - Has the customer had any issues/problems? - Would this customer consider the use of SAN's a success? We received many responses. Our thanks to all who answered! The results, as compiled by Jim Troisi, of IBM, follow my signature. Best regards, Fred Hubbard IIUG -o- Many customers are using SANs: [Customer and contact names deleted until legal releases are obtained] Customers were running IDS 7, IDS 9 and XPS. I did not notice a correlation between this data point and any of the other data collected. Customers had SANs on all supported XPS platforms: AIX, Compaq, HP, SUN. Customers were storing on SANs anywhere from 100GB to 116 TB (yes 116 TB!) of data. Not all of this data was in the Informix databases. The XPS systems tended to have larger amounts of data than the IDS systems, but not always. Customers were using IBM, EMC, HP, Hitachi, NetApp, and Compaq SAN systems. There was a small correlation between Sun HW systems using EMC SANs and AIX HW systems using AIX/Hitachi SANs. Most customer had no problems at all using the SAN systems. A few customers mentioned minor problem. Interesting quotes included: - Poor planning/implementation on <the customer's> part (lead to performance and HA problems). - At <customer>, we have been using SAN disk arrays for several years. Years ago, we adamantly refused to use RAID5 disk arrays because of the write penalty and we insisted on hardware mirrored disks. With today's huge write caches and much faster disks, we have no problem using RAID5 on SAN arrays. (Editor's note: This was echoed by other respondents.) - Using the SAN also allows us to effectively use HP's ServiceGuard and Sun's Veritas failover packages to move instances and all their associated application processes from one box to another in about four minutes. - The use of SAN has not really affected the way we design our databases. When we first started using the SAN, we thought it would be a great idea to make one huge dbspace for all of our user created databases. This made monitoring available space easier as well as deciding how to lay out the tables. Of course this eliminated the parallelism available from dividing your data across multiple dbspaces. Now we have gone back to using multiple dbspaces and our archives, queries, etc. are running much faster. - From an issues perspective, I believe the EMC arrays to be generally more reliable than any other non-SAN solutions that I have worked with. (Editor's note: Many people asked me which was the best SAN vendor to buy. This comment is one person's unsolicited comment. I do not believe the database group should be suggesting one SAN over another. Now, (IBM's) SHARK group might have a different take on this.) - I can categorically state that we have had the most success when we have treated our SAN solutions in the same way that we would treat non-SAN disks; data placement, physical disk (spindle) separation of data, etc. are all important factors in maintaining good performance from these vendors. The customer has written two white papers on the customer's installation, one on the migration itself and the problems it solved, the other on the integration of EMC, Omniback and IDS onbar to reduce restore times down from 36 hours to 1 hour. --Jim Troisi IBM Data Management, Senior Software Manager