Unsolved
This post is more than 5 years old
34 Posts
0
5398
Symmetrix VMAX & VMAX3 Logical Structure
Hello,
Basically I wanted to confirm my understanding of the logical structure within the Symmetrix VMAX and VMAX storage arrays.
Correct me if I'm wrong, I'll go top to bottom.
1) Disk Group - A group of identical disks in terms of technology, size, speed etc. *VMAX All-Flash has only one Disk Group.
2) RAID Group - A group of disks from the disk group that have the same RAID protection. *Question: Can I have more than on RG per DG or is it always 1:1 ?
3) Data Device (TDAT) - A logical slice of a physical disk in a particular RAID Group.
4) TDEV - A logical slice of a TDAT, smallest logical device in the storage array. *Question: I've read that TDEVs are comprised of Hypers which are slices of a physical disk. How does that work exactly?
Now, regarding provisioning and FAST and VP Tiers:
-A meta device is comprised of several TDEVs and provisioned to a host as a LUN. *VMAX3 does not have metas.
-A thin pool is comprised of Data Devices (TDATs). *Question: Can there be several pools per RAID GROUP or is it always 1:1?
-Are VP Tiers based on Thin Pools, meaning every thin pool is a separate VP Tier and FAST moves data between them?
Summary of all questions just in case:
1) Can I have more than on RG per DG or is it always 1:1?
2) I've read that TDEVs are comprised of Hypers which are slices of a physical disk. How does that work exactly and what are Hypers exactly in technical terms (I've seen several different explanations)?
3) Can there be several thin pools per RAID GROUP or is it always 1:1?
4) Are VP Tiers based on Thin Pools, meaning every thin pool is a separate VP Tier and FAST moves data between them?
PedalHarder
465 Posts
1
September 23rd, 2018 19:00
For 1) It is a 1:many relationship between a disk group and RG. A disk group consists of one or more RAID groups.
2) you need to logically carve up physical disks. If you didn't, you would end up with a 1:1 relationship between the TDAT and the RG. This is some history here. A physical disk could contain different geometries. i.e. FBA for Open systems and CKD for mainframe. To do this you need a logical construct over the physical disk. We also create multiple TDATS on the disk for performance purposes.
3) Yes a RG can contain have TDATS that belong to different pools.
4) VP tiers consist of disk groups.. As such a VP tier may contain multiple pools. See the symtier command.
LearnEverything
34 Posts
0
September 25th, 2018 06:00
Hello Jason, thank you very much for answering.
1) Could you give me a specific scenario where I would want to have several RGs per DG?
2) I do understand the principle behind that, however my question was more specific to what "Hypers" or "Extents" (seen both terms used interchangeably) actually mean? Do they mean a specific amount of blocks, size etc.? Because I have seen both terms being used for different sizes which is why I got confused and wanted to clearly understand what the terms mean.
3) The nature of my question would be the same as before, why would I have several pools across the a specific RAID Group instead of just having one pool?
4) Got it, thanks!
PedalHarder
465 Posts
0
September 26th, 2018 16:00
A single RAID group can only contain a specific number of disks, based on the RAID protection. 14+2 is the largest VMAX has ever used (16 disks). Of course systems are much larger than 14 drives so a Disk group will consist of as many RAID groups to meet the capacity requirement.
Hyper and extent are different terms. Hyper is a logical construct on the physical disk. TDATS are made from Hypers in a Thin array. Extents are a logical construct of a thin pool. In a VMAX 3 the extent size is 128KB.
A customer may choose to split a DG up into more than one pool for their own reasons. I have seen a separate clone target pool for example.
LearnEverything
34 Posts
0
October 1st, 2018 05:00
Great! I understood everything except for Hypers/Extents.
From what I understand from you, a "Hyper" is a logical volume that sits on one physical disk. Several Hypers on the same RAID Group are joined to create a TDATs which in turn create a thin pool? I was under the impression that TDATs are made up of TDEVs, is that not the case?
"Extents are a logical construct of a thin pool" - what does that mean exactly? As I said earlier, I was under the impression that TDEVs are the smallest building block. Could you explain what they are in relation to the physical disks and then to the logical volumes (such as TDEV)?
PedalHarder
465 Posts
0
October 11th, 2018 19:00
Extents are not exposed to the customer and there is no need to use or know about them from a storage provisioning / management perspective. The array uses them internally. For example a FAST movement is performed at the extent level.
LearnEverything
34 Posts
0
October 14th, 2018 10:00
That's some great insight, thanks!
Now regarding Hypers - I understand they are logical slices that make up the thin pool, what I don't understand then is what the purpose of TDATs are? I mean why have Hypers if you have TDATs?
PedalHarder
465 Posts
0
October 14th, 2018 21:00
You need a layer of virtualization / abstraction between the disk and the host device to do thin provisioning. Thick provisioning presented the hypers to the hosts. But to get the benefits of pools and thin provisioning the back-end is virtualized with TDATS.