Data Progression Question

Couple of things here...

First - its important to know which Data Progression we are talking about - if the system is a hybrid flash, it works differently then a spindle system. From your question, I am going to assume spindle since you are providing the time frames. 

Second - Storage Profile determines which tiers of disks are available for write data (non replay) data and read (replay) data. Moving things up and down in the Tiers depends first on which Tiers are available for use. 

Third - In the SC series pages are on of a few types - this matters in the algorithm. Accessible or Historical AND Recently Accessed, Not Recently Accessed, or Non Accessible.  So a non-accessible block (kept for replay consistency) immediately moves to lowest tier and stays there. Where as the other types can move around per the timing algorithm. 

I am looking for my most recent "life of a page" presentation to see if it describes whether the threshold is met and then acted on or if its actually on the day of - this has varied by firmware release over the years - but generally its threshold then next progression window moves the block. Tier1 would only be used if its selected in the storage profile - but never back to RAID10 as that would not have any benefit.

Yeah with an all flash array, it does everything at time of replay, because flash is just much faster in general.

The storage profile is all 3 tiers, ssd t1, 15k t2, 7k t3

So "this has varied by firmware release over the years - but generally its threshold then next progression window moves the block." From this I read it as on the 12th day it will start moving it down? So if the data was in t1, it would start to move to t2 and depending on the type of page it would either go directly to t3, or go to t2 and so on?

Thanks for taking the time to respond

Arrays that contain WI and RI flash operate a little differently. It doesnt have anything to do with the speed of the flash, just the idea of wanting to clear the WI space ASAP and so a simple RAID10 -> RAID5 T1 -> T2 progression runs much more frequently (hourlyish). 

If you have all three tiers marked on both the Replay and Write side, then a highly used block could be moved all the way up to Tier1. 

Like anything - it depends on the timing of things. If you write a block at 6p (with a DP starting at 7p) and wait 12 days, then it will start to move an hour later. If you write a block at 8p then it will probably be batched the next day. There are exceptions to this example in both directions. 

"Arrays that contain WI and RI flash operate a little differently. It doesnt have anything to do with the speed of the flash, just the idea of wanting to clear the WI space ASAP and so a simple RAID10 -> RAID5 T1 -> T2 progression runs much more frequently (hourlyish)"

yeah sorry this is what I meant. It will move data out of RADI10 in T1 on demand, esp when that bit is full to free up space, it can do this because of the speed/IOPs of the WI SSDs. it wouldn't be able to do this with spinning disks. Well thats my understanding anyway!

"If you write a block at 6p (with a DP starting at 7p) and wait 12 days, then it will start to move an hour later. If you write a block at 8p then it will probably be batched the next day. There are exceptions to this example in both directions. " Ok from this I read it as:

Data written at 6pm , DP runs at 7PM, so after being written  at 6pm on day 1 on the 12th day @7pm it will move the data down, and depending on the page it may go to t2 or directly to t3.

If the data was written @8pm, it would move down on the 13th day @7pm when DP runs.