This post is more than 5 years old
125 Posts
0
2140
December 6th, 2010 13:00
Can concatenated Meta be of different hyper sizes ?
Hi all,
I am stuck in a situation where i need to create LUNs of different sizes, I have hyper sizes of 16,33,67,134,240 plus something...what is best practice to create LUNs of sizes like e.g..350gig..considering best practices and low wastage on space... ??..This is on VMAX and thin provisioning in place..
Here questions like many meta members, few meta members, number of hypers in a meta should be even or multiple of 2 (2^n).
No Events found!
emcsanadmin
54 Posts
0
December 6th, 2010 18:00
SAN_Ak,
in our environment we have VMAX & we do thin provisioing on it.
for your requirement 350G we do in this way.
we will create 5 devices of size 70G once devices are created i will form them as Meta.
once meta is created i wiil bind the meta head to thin pool( FC or sata) based on the rek.
once it is binded to the pool i will assign to the Storage group which is comprise of
storage grp = portgrp+ masking view+ initiator grp & go to the masking view to see the LUN id for the device i created.
NOTE: in our environment we will create the deivces based on our requirement & more over we can create a single TDEV or hyper of size 240G on the vmax.
hope this helps....
SAN_AK
125 Posts
0
December 6th, 2010 14:00
its symmetrix VMAX model, Enginuity 5874.
xe2sdc
2 Intern
•
2.8K Posts
1
December 6th, 2010 14:00
You can choose whatever number of members.
EMC suggests an even number of members. Better if it's something like 2^n. But you CAN use odd number of members, if you need.
With an odd number of members (or a number that don't fall into the 2^n suggestion) it's difficult spreading the workload evenly in the backend. But usually this isn't a real issue.
If you are using RAID in the backend, don't spend too much time in choosing the number of members.
If you are using mirroring, try to use 2^n rule but don't waste too much space (and time) in choosing the number of members :-)
xe2sdc
2 Intern
•
2.8K Posts
1
December 6th, 2010 14:00
Not sure... Pls wait replies from other contributors...
You talk about concat meta. AFAIK concat isn't available in recent codes. What code version are you running in your storage? What storage are you using?
SAN_AK
125 Posts
0
December 6th, 2010 14:00
Thanks, can i mix different types of hyper sizes as members with in same meta ?
xe2sdc
2 Intern
•
2.8K Posts
0
December 6th, 2010 15:00
We definetly need an answer from someone with vmax skills... :-)
Can't answer on vmax, yet. :-)
Boom1
131 Posts
1
December 6th, 2010 17:00
yes.. you can..
Meta Device Members (3) :
{
----------------------------------------------------------------------
BCV DATA RDF DATA
---------------------------- --------------------------
Sym Cap Std Inv BCV Inv Pair R1 Inv R2 Inv Pair
Dev (MB) Tracks Tracks State Tracks Tracks State
----------------------------------------------------------------------
--> 0xxx 34856 - - N/A - - N/A
0xxx 34856 - - N/A - - N/A
0xxx 1688 - - N/A - - N/A
----------------------------------------------------------------------
71400 - - - -
}
You need to make sure, all devices are unbind from pool....
SAN_AK
125 Posts
0
December 6th, 2010 17:00
Can you please guide me which one is the best practice for creating 350 gig LUN for windows host. with the device standards mentioned above.
Do you see any other possibilties, or any standard method for dealing this situations as you have somany combinations of summing up devices to 350gig.
PedalHarder
3 Apprentice
•
465 Posts
0
December 6th, 2010 18:00
You mention in your original post that virtual provisioning is in place. With VP, the front end meta devices are assigned to a device pool rather than a tied directly to back-end hyper volumes. so with VP in place, the practice of having uniform meta sizes can be relaxed somewhat.
Rather than pre-creating the TDEVS at standard size, if it is going to work better for you perhaps you could create your TDEVS as part of the storage allocation process with custom meta member sizes.
Here are some considerations with the meta member size:
- SRDF/S can have one outstanding write per device per path, so a single large device will potentially have limited write rate compared to a meta device.
- Symmetrix devices numbers are finite resources, there are limits on the number that can be presented to an FA port, and an overall limit. So you need to be able to address all your back-end storage while keeping within the limits of your available Symmetrix resources.
SAN_AK
125 Posts
0
December 6th, 2010 20:00
here I have 67.4GB as standard in my environment, 5 devices cannot makeup 350 that way, All I am trying to know that i can mix up different device sizes in a meta, that got answered in this post. I can understand if you have 70 gig dev that could make up simply 70*5 350gig there.