ITEM: I6561L

Questions on 7135 RAID

Question:

Will have apllications runnng on RS/6000 and AIX 3.2.5. Considering 
RAID-0 and RAID-5 technologies. Need to understand the following as 
related to RAID-5: 

1. If a disk module goes bad, can I continue operations while the 
   rebuilding process goes on. Specifically, if I request data from the 
   bad disk, does the system dynamically rebuild that access or must I wait 
   until the rebuild of that disk is completed?

2. If I can continue operations during the rebuild process, what is the 
   performance degredation during the rebuild.

3. Lastly, how long does the rebuild process take. Are there estimates based 
   upon file size? 

Response:

In RAID 0 - if a disk goes bad, all data is lost, must fix hardware
then restore data from tape.

In RAID 5:

1. Operations continue if only 1 disk goes bad.  After replacement of
bad disk, rebuild occurs during regular operations.

2. Performance is degraded.  When a disk is lost, read operations now
require reading all disks if the data was on the lost disk. 
Thus the data is recalculated from the remaining data and parity date.
When a rebuild occurs, both rebuilding and regular I/O can
occur, but rebuilding degrades performance also.

3. Rebuilding varies depending upon user definable parameters (the
"reconstruction frequency" and the "reconstruction amount in blocks") in 
READI/REACT.  Time for completion is based upon the disk size, not file sizes.
Reconstruction times can vary from less than 3 to more than 20 min/GB.
A moderate reconstruction rate may take about 7 min/GB.  Thus if he
has 2GB disks, it will take about 14 minutes on the 7135.




Support Line: Questions on 7135 RAID ITEM: I6561L

Dated: April 1994 Category: N/A

This HTML file was generated 99/06/24~13:30:44

Comments or suggestions?
Contact us