DMX hyper size

most customers have some sort of standard building block in such a way that there's very little to none capacity loss. for example: when you have 146GB disks in Raid5(3+1) and you cut it in such a way that you have 8.51GB symdevs, each hyper is 2.836GB, so on each 146GB disk are roughly 146/2.836 = 51 hypers / slices per physical disk. 51 x 2.836 = 144.6GB, so you loose 146 - 144.6GB = 1.4GB per physical disk.

Standard thick provisioning or VP (thin) provisioning?

In general, fewer larger hypers (within reason) gives better overall system performance than lots of smaller hypers on the disks.

Yeah, you are true. 

Its for standard thick provisioning.

Thank you RRR.

I have also different size of hypers in my Symm envoronment starting from 2gb. I got this question during an interview. So.. I thought some standard hyper size defined by EMC. LOL

Thank you

Sanjit Achary

As per you update, the hyper count will differ depends on the size of the disk and depends on the size of each lun.. Please confirm this

Symcfg list –sid -v

Correct.

If you have 300GB disks and still use the 8.51GB symdevs, you can "carve" more hypers or slices from each spindel. If the symdev size grows, you need bigger hypers (not something you can configure by the way, since you can only enter the symdev size, not the hyper size). Bigger hypers mean that you will get less hypers from each spindel. In the old days with DMX3/4 best practice was to stay below 64 hypers per spindel, but I'm not sure this is still valid.

The max number of hypers per spindel is a system wide variable, so if you need to set it to 128, because you decide to have 8.51GB symdevs on 300GB disks, you can also have 128 on other disks. Less slices per spindel is better for performance.

the actual size of a LUN is determinded by what you eventually map and mask to a host. A single LUN can be composed from several symdevs, each of which is composed from several hypers. A Hyper or slice is a "physical slice" taken from a physical disk !

So what everyone calls a LUN, is what the hosts / servers eventually sees on their HBAs. what's behind a LUN, is a symdev or a META (several symdevs bound together).

Is there any CLI command to check the "max number of hypers per spindel" in a particular DMX?

Welcome, Mr Dynamox !! I'm glad you could finally join us

it's your favorite topic so i stay out your way

hahahaha, I'm glad if you can confirm everything I've said so far