I plan to migrate our V9 DB from AIX server to OPE10.1B on Win 2K3 platform.
I read in the OE documentation that Progress recommends use of RAID10. I read also the importance of the data layout and specially the distribution of the data across the storage areas.
I would like to know if this data layout is so important in case of RAID10 implementation. Because the data are stripped through several disks at OS level (2x4 disks in my case), I don't see how to affect physically the extents to the disks, specially for index areas an primary recovery areas. I think this doesn't make sense.
Could it be better in performance to have a RAID10 volume for the data and a RAID1 for the indexes and BI?
What is your opinion about that?
Thank you in advance for your help.
There are lots of choices possible and many variations in requirements as well. I don't recommend putting indexes in raid 1 (mirrored) and data in raid 10 (mirrored and striped). Put both in raid 10.
What works well in many cases is:
- raid 10 for data and index extents so disk load is evenly distributed across the drives, with as many drives as you can get.
- before image log on a separate drive if possible. If not, put before image log on same stripe set as the data extents.
- after image log should go on physical storage /seaprate/ from data and before image laod. Do this so if the array the data is on gets trashed, you don't also lose that after image logs.
Above assumes you have one database. If you have more, you probably can't have a bi disk for each database, so in that case, you are better off with bi logs on same stripe set as data and index extents.
Message was edited by:
Thank you for this answer. Effectively we have many databases and we can't have one BI disk for each one.
Our application is MfgPro with about 190 users in multiple environments: 6 database sets.
Typically, we have no many transactions in this application and no performance issues on writting mode. I experienced the highest load with this application in reporting mode with 3 or 4 big tables (more than 2 Gb).
In this context, I ask me also if it could be better to group these big databases on a separate disk array. What do you think about that?