Hi all,

I'm looking for an mathematical or physical explantation about the Chunksize of 42MB which is used in Hitachi Storageboxes.

Can anyone help with a few slides, or an paper ?

thanks in advance and best regards

Ralf Siekierski

Hi all,

I'm looking for an mathematical or physical explantation about the Chunksize of 42MB which is used in Hitachi Storageboxes.

Can anyone help with a few slides, or an paper ?

thanks in advance and best regards

Ralf Siekierski

- 1 person found this helpful
I think that piece does answer why 42MB page size doesn't it?

"Now, unless its changed, this 42MB is taken from contiguous stripes from a single LDEV in a single Array Group. This basically means that each of these 42MB pages is allocated from a single Array Group (8 disks). Subsequent pages allocated to the LUN will

*probably*be allocated from another Array Group in the Pool….. So over time your LUN should have its blocks nicely balanced across all spindles in the Pool. A form of wide striping."and

"Now for the

*slightly*more interesting part (I stress the slightly). You will find that every one of the stripe sizes listed above divides perfectly into 42MB. If you dig further, you will also find that 42MB is the lowest number that all of the above stripe sizes divide perfectly into, without generating a remainder. " - 3 people found this helpful
Hi Ralf!

Actually 42MB (and it is page, not chunk) is the least common multiple for stripes of all RAID levels available in system.

Stripe unit size is the minimum data block on the individual drive in the RAID. Actually HDS calls them chunks, but this might be confusing with other chunks in their terminology. Stripe unit size in HDS systems is 512KB. Another important thing to know is stripe that is the sum of data blocks across the set of drives in row. Thus assuming we count the blocks on Data (D) disks we get:

RAID10 (2D+2D') - stripe is 1024 KB (1 MB), D' means copy

RAID4 (3D+1P) - stripe is 1536 KB (1.5 MB)

RAID5 (7D+1P) - stripe is 3584 KB (3.5 MB)

RAID6 (6D+2P) - stripe is 3072 KB (3 MB)

RAID6 (14D+2P) - stripe is 7168 KB (7 MB)

And as you can see we can place 42MB on all RAIDs within integer number of rows - 42 with 1 MB stripe, 28 with 1.5 MB, 12 with 3.5MB and so on.

I think the 42MB with the least common multiple etc is just a guess work. That is not the answer.

If that was the case then it would have been true for mainframe hdp as well.

I think the correct answer is on how much area is there in the shared memory for the hdp. The

shared memory size decides on what should be the page size. And with the first implementation

of HDP on USPV that may have been the reason. Since then it has continued that way.

- 2 people found this helpful
Hello Bhaskar!

I could simply reply that 42 is the answer to "The Ultimate Question of Life, the Universe, and Everything". That is why HDS chose it )))

But with the following you can find mainframe comments and calculations.

For mainframe environment HDS array uses 58 KB track (tracks and cylinders are still used for internal operations). Eight adjacent tracks are taken from the same drive before switching to the next drive in the row of the parity group. Thus we have 464 KB as the size of stripe unit. Then go all the stripes:

RAID10 (2D+2D') - stripe is 928 KB

RAID4 (3D+1P) - stripe is 1392 KB

RAID5 (7D+1P) - stripe is 3248 KB

RAID6 (6D+2P) - stripe is 2784 KB

RAID6 (14D+2P) - stripe is 6496 KB

The least common multiple for stripes is 19488 KB but HDS doubles it to 38976 KB (672 tracks of 58 KB) for efficiency (smaller page size requires more metadata in Shared Memory). To be honest, HDS talking about 38 MB page for Mainframe DP but it's not exactly that (38 MB = 38912 KB).

Hope this helps.

Dynamic Provisioning: The 42MB page unravelled – nigelpoulton.com