First page Back Continue Last page Graphics
The S.A.M.E. Methodology
“Close to optimal [disk] performance for all workloads”
Stripe all Oracle files across large number of disks
Place active files on outer half of disks
Mirror for redundancy
- S.A.M.E. architecture
- RAID 5 not best performer
Use volume manager to implement striping and mirroring
Notes:
Though the cost of very large disk devices (e.g. 80GB) is always dropping, and thus it is easy to place more of your data on fewer disks, this is not recommended for best overall Oracle performance. The Oracle benchmark referenced earlier in this lesson showed that the “stripe and mirror everything” (S.A.M.E.) configuration provided the best overall performance. The white paper states “The SAME configuration produces close to optimal performance for all workloads: OLTP, Warehouse and Batch.”
The S.A.M.E. methodology calls for:
Striping all Oracle files over many disks (the more the better) and using a 1MB stripe size. Note that the 1MB stripe size has to do with disk transfer speed and seek rate and could change as these characteristics change with advancement in disk technology.
Placing high use application files and system files (e.g. redo logs) on the outer half of the disk device (because the transfer rate on the outer half of disk devices is faster than the inner half).
Using disk mirroring rather than RAID-5 (for availability) because RAID-5 introduces unnecessary overhead.
Notes for this slide continue on the next page…