Hi,

The latest FAST VP paper has been published, but it is still a little confusing as to how the polices work:

Highest tier

Initial placement should nearly always be on the highest tier as 10% of capacity is always kept free

The coldest slices will be moved to lower tiers based on the defined schedule, but only at the expense of other Highest tier LUN slices

Hot slices on lower tiers will be moved up based on the defined schedule

If the size of the slices allocated to all highest tier LUNs is greater than the tier size, then no other tier priorities will ever have their slices on the highest tier - is this correct?

Lowest tier

Initial placement should always be on the lowest tier

The slices will never move off the lowest tier even if they are very hot - is this correct?

Auto-tier

Initial placement will be to spread all slices across all the tiers based on free capacity

The slices will never move to the highest tier if it is full with slices from highest tier priority LUNs even if the IO is much higher on the slice than those already on the highest tier - is this correct?

The slices will moved to the appropriate tier based on activity (i.e. typically T2 or T3)

So if you have:

1. A LUN that needs best performance - set to Highest tier

2. A LUN that does not need good performance - set to Lowest tier

3. A LUN that needs good performance - set to Auto-tier

This does mean that you could have very cold slices of the highest tier sitting on T1 or T2 storage, very hot slices of auto-tier sitting on T2 and very hot slices of lowest tier sitting on T3 - is this correct?

I am not 100% sure about this, but it is my interpretation of the paper.

Any comments would be much appreciated.

Many thanks

Mark