Storage XP512

[Mark Huff II]

Would there be a performance difference between creating 90 36gb logical volumes vs 1 3.24tb logical volume? (these are 36gb chunks)

If a lun fails do I lose the entire logical volume or do I just lose the information on that lun?(This is raid 5)

Making one drive simplifies our exporting and mappings.

Thanks!

[Vincent Fleming · Accepted solution]

If configured properly, there is very little inherent overhead to LUSE (combining LDEVs into a single, larger LUN). It's less than a 5% difference.

The trick is to be sure to configure it properly - which means picking source LDEVs that are spread accross as many array groups as possible, and where they are in the same array group, they should be consecutive LDEVs (next to each other).

This will provide performance via load balancing the drives, as well as reducing head thrashing a bit.

Here's the caveat - LUSE concatenates the LUNs... ie: there's no striping. Striping from a host is generally faster. Note that striping will not always increase your performance, though - so it's hard to say how much less (if any) your performance will be without testing it in your actual environment.

I would not worry about a LUN failing - in either configuration (host or array-based concatenation/striping), if a LUN fails, you lose it all. BUT - I haven't seen a LUN fail on an XP, ever.

Have faith, and take backups just in case. It's a very fault-tolerant machine. It's so fault tolerant that we don't even talk about how it protects against data loss - we talk about how it protects against loss of access! (a LUN going offline for a while without data loss) There are no worries about data loss anymore.

Good luck!