I am in the process of designing the migration of an existing DB2 production database to new storage.

Most likely we will be using RAID 10. By way of example let's say my two choices are:

1. Single RAID 10 array with 8 primary disks. DB2 tablespaces consisting of a single container each.

2. Two RAID 10 arrays with 4 primary disks each. DB2 tablespaces consisting of two containers each, one on each array.

From a DB2 performance perspective, is there any difference between these two approaches?

It's unlikely I will have time to benchmark these two options and so any advice to get me going on the right track would be much appreciated.

Thanks in advance.

Hello,

On 17.05.12 4:31 , Desmodromic wrote:
Hmmm, as with most theoretical questions, the answer is: it depends.

There are several levels and forms of striping:

1) storage subsystem
2) OS - LVM
3) DB2

The best for performance is to use only 2 out of the 3 options. I usually tend
to go with 1 and 3 (that is, if you are the storage admin as well).

It also depends on the workload and how many IOPS your workload requires.

The best is to use as many disks as possible in an array. You can still create
2 LVs per tablespace on the storage.

It also depends on the connectivity to the storage. How many fibre channel
adapters, how many GB cache per adapter, ...

--
Helmut K. C. Tessarek
DB2 Performance and Development
IBM Toronto Lab