Answers to Chapter Review Questions

     
A1:

RAID 0 provides striping across multiple disk drives . Where an application performs serial IO, a RAID 0 volume can offer significant IO performance benefits. By definition, random IO means that the actual IO performed may not be distributed over all disks in the stripe set. It is feasible that, purely by chance, the random IO is actually performed by a small subset of the disks in the stripe set, even to the extent of only one disk performing all the application IO. It is the random nature of the IO that makes determining the benefits of RAID 0 more difficult to quantify. Hopefully, over time the randomness of the IO will eventually mean that all disks in the stripe set perform a proportion of the application IO. In this way, we will hopefully see an overall improvement in IO performance.

A2:

RAID solutions with a single parity disk offer high availability solutions but are seen to be poor from a performance perspective. A single parity disk will become an IO bottleneck when data in the RAID volume is changing frequently. When a data item changes (be it a block or a byte), the parity data needs to be read from the parity disk, the new parity data calculated from the new data, and then the new parity data written to disk. This is known as the read-modify-write performance problem attributed to any RAID solution that utilizes parity data. With RAID 5, the "parity IO" is spread over all disks in the RAID volume, whereas with single-disk parity solutions, the "parity IO" is concentrated on one disk, which becomes the IO bottleneck for the entire RAID volume.

A3:

Parity

1

1

1

1


A4:

RAID 3 utilizes byte-level striping. Operating Systems and applications deal with disk IO in blocks (of varying sizes). The size of a RAID 3 stripe is sometimes calculated by dividing the size of an OS/application block by the number of disks in the stripe set. Even when this is not the case, an OS/application "block" will be spread over multiple disks in the RAID 3 stripe set. When an OS/application "block" is modified, multiple disks need to perform an IO in order to read the entire OS/application block into memory. Once all IO has been performed, the new parity information can be calculated. When a significant number of OS/application "blocks" are being modified, this generates a significant amount of IO. Compared to RAID 4 where a single OS/application "block" could be satisfied with an IO to a single data disk + IO to the parity disk, this offers significantly less numbers of IO operations and, hence, can offer significantly better performance. However, it should be noted that having a single parity disk would become, in itself, an IO bottleneck when data is being modified constantly.



HP-UX CSE(c) Official Study Guide and Desk Reference
HP-UX CSE(c) Official Study Guide and Desk Reference
ISBN: N/A
EAN: N/A
Year: 2006
Pages: 434

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net