Ask the Expert

Migrating data from four 400 GB SATA drives

I currently run a system using four 400 GB SATA drives configured with RAID 5. I want to replace all drives with four 750 GB drives. Can I do this by swapping individual drives, after each is rebuilt? The SATA controller is the NVIDIA MediaShield, and supports RAID 0, 1, 0+1, 5 and JBOD.

Continue Reading This Article

Enjoy this article as well as all of our content, including E-Guides, news, tips and more.

Tragically, your proposed method is not the best way to migrate your data. Remember, RAID rebuilds take quite a long time and are a resource-intensive operation. So expect your application to nicely suffocate while this is happening. RAID rebuilds are a protection mechanism and are not designed for data migration. To this point, assuming your 750 GB drives are supported with your RAID adapter, the RAID group will only utilize the first 400 GB of storage on each 750 GB drive.

Remember, RAID groups typically require that all the drives in the group are of the same type, speed and size (see my previous articles for more details on this). This means that even after you go through your "migration," you'll still only have four 400 GB SATA drives worth of available storage. The remainder will be inaccessible.

The usual way to migrate data with internal server based storage is to copy the data to another separate set of drives or temporary storage, internally or on your network or a *wince* large USB HDD (the *wince* means "not recommended but can suffice if you're in a pinch"), install the new drives and copy everything back, remember to preserve all file attributes such as permissions and access dates, etc. You can use a tool such as Robocopy.

It's headaches like these where a consolidated mid-tier storage solution such as those from EMC Corp., IBM and Hewlett Packard Co. comes in handy. They can migrate data from one RAID group to another without any disruption and usually while the application is still accessing the volume! If your company can afford it and has a significant enough storage management requirement, I strongly suggest investigating this.

This was first published in January 2008