ASM Even Read

Question

Hi,
My understanding is that in Normal/High redundancy disk groups the Even Read feature ensures that a new I/O request is sent to a disk with the least amount of outstanding I/Os.
But is the amount of outstanding I/Os calculated at each ASM instance separately or across all ASM instances collectively?
This would make difference in some configurations with asymmetric access to storage between nodes where storage network is the bottleneck and not the disks (and where Preferred Read cannot be fully used).

Answer 1

> Even Read helps in scenarios when distribution of primary copies across failure groups becomes uneven, and as a result distribution of read I/O would becomes uneven too.
An example:
One disk fails, now its partner disks have to serve reads from the secondary extents that had primary copies on the failed disks. But they also have to serve reads from their own primary extents. So, some disks may become more loaded than some other disks. Even Read eliminates this imbalance by keeping stats of how many I/Os are served by each disk.
Again, if someone has a more accurate description please share.