Change raid1+0 to autoraid on va7100

[Jimmy_13]

Can i increase the capacity for data in the va7100 when change form raid1+0 to autoraid?
Can i do on-line?
In this moment i have 14 disks of 36 GB, 467 of total capacity but there are only 237 GB for use by the redundancy.
I supose that to change of raid, i have more than 237 GB for use and the LUNs and data won??t missed

[Eugeny Brychkov]

Jimmy,
from the manual (available on the web):
- Changing from RAID 1+0 to AutoRAID. The RAID level can be changed from RAID 1+0 to AutoRAID on-line. However, it is recommended that you backup all data on the array before changing the RAID level.
- Changing from AutoRAID to RAID 1+0. The RAID level cannot be changed from AutoRAID to RAID 1+0 on-line. This change requires a complete reformat of the entire array, which will destroy all data on the array. Before changing from AutoRAID to RAID 1+0, backup all data on the array for restoration after the format and RAID change are complete
Eugeny

[Brian M Rawlings]

Jimmy: You will get better space utilization with AutoRAID mode. Even in AutoRAID mode, the array tries to store all data as mirrored, but once you pass 50% utilization (too much for mirroring of the data), the array will begin to migrate old unused data to RAID-5, freeing up additional space.

In the worst case, the array will store 90% of the data as RAID-5, keeping 10% of the space available for RAID-0/1 mirrored storage (for the most active data that needs the fastest access). The migration is all based on time stamps of last access to any block, so this may or may not be a good algorithm for your situation.

Say you actively use, say, 50% of your data, all the time, maybe you have an active database there. The rest is archival files, seldom touched.

Then, what you would see is that, as you keep putting more data in, and some of your active blocks begin to get migrated to RAID-5, performance will begin to lag.

In order to see where the "knee of the curve" is, for your particular data access pattern and type, you should carefully monitor disk access statistics and overall performance.

It will go along OK as you add data to your "available space", then, after some data space growth, you will see a more significant performance hit, and it will stay and not be transient. That is your particular pain point, and you should look at adding more disk drives/trays, etc, to get back more mirrored space for your busy blocks.

This sounds like a fairly crude and rude mechanism for capacity planning, and it is. There are much more exhaustive analysis techniques that *may* predict when you will need more storage. They are not generally easy nor free. Some people need them, some get by with the quick and dirty method.

The reason the simple way works is, you don't suddenly stop working, you just see access times begin to go up, and you know it's time to go spend some money. Nothing broke, and if you are the one watching it the closest, nobody is suddenly climbing your tree for response time. Now you have a few weeks or a month to get some more space, which may or may not be enough time for you and your company.

On the other hand, lots of people out there will be happy to con$ult with you about trend analysis and capacity planning... if that's what you need, that's what you'll have to pay for.

Regards, --bmr