ASM/MetaLUN

I have been reading through some EMC docs as well as Oracle ASM stuff, trying to understand what would be a reasonable setup (the recent Storage Array thread in this list also helped). My understanding so far is that increasing the number of disks in a RAID group is good because that spreads I/O over more disks. The disadvantage is that having more disk also increases your chances of disk (or multiple disk) failures. Also, performance of a RAID5 disk group after losing a disk is worse as the disk group size increases (you need to read across all disks in the stripe to reassemble the data). One way around this issue is to stripe across disk groups--many disks, just a few disks per disk group. This can be done at the hardware level (i.e. emc's metaLUNs. If you add more iron and then stripe across MetaLUNs does that become a 'Heavy MetaLUN'? My face is turning a shade of deep purple with embarassment after writing that). Oracle's ASM in 10g will also let you stripe across disk groups. Just wondering what the advantages/disadvantages are of these two methods. I know ASM is new technology and hence still a bit suspect, but since it is implemented by the database it is supposed to evenly spread the datafiles across the disk groups. Is there a difference in stability, performance, monitoring capabilities, ease of adding/removing a disk group? By the way, I am assuming that when blocks are written to, they would be distributed evenly over all disks. Anyone know if that is how the algorithm really works?

Thanks.

Henry

--
http://www.freelists.org/webpage/oracle-l