VMware Cloud Community
aleceiffel
Contributor
Contributor
Jump to solution

LUN sizes

I am setting up LUNs on a Lefthand NSM 2120 G2 iSCSI SAN device for use with VMware. What are the recomended LUN size limits for a setup like this?

0 Kudos
1 Solution

Accepted Solutions
Rockapot
Expert
Expert
Jump to solution

I usually recomend clients allocate LUN's of between 500Gb to 800Gb however it really depends on sizing of your environment. 500 - 800Gb LUN's are fairly simple to manage. Some clients tend to go with 1 LUN per VM however I strongly discourage this as it is a management nightmare.

If you are performing a migration from an existing physical environment or from an existing virtual environment it is important to design to storage correctly from the start.

Greenfield sites are a little easier to design as there is more room for change.

Do you have an existing VM environment you are migrating or is this your companies first VMware implementation?

Carl

View solution in original post

0 Kudos
8 Replies
lamw
Community Manager
Community Manager
Jump to solution

Take a look at this post: http://itknowledgeexchange.techtarget.com/virtualization-pro/choosing-a-block-size-when-creating-vmf...

Also take a look take a look at Duncan's post here too for some advice: http://www.yellow-bricks.com/2009/03/24/an-8mb-vmfs-blocksize-doesnt-increase-performance/

=========================================================================

William Lam

VMware vExpert 2009

VMware ESX/ESXi scripts and resources at:

If you find this information useful, please award points for "correct" or "helpful".

0 Kudos
Rockapot
Expert
Expert
Jump to solution

I usually recomend clients allocate LUN's of between 500Gb to 800Gb however it really depends on sizing of your environment. 500 - 800Gb LUN's are fairly simple to manage. Some clients tend to go with 1 LUN per VM however I strongly discourage this as it is a management nightmare.

If you are performing a migration from an existing physical environment or from an existing virtual environment it is important to design to storage correctly from the start.

Greenfield sites are a little easier to design as there is more room for change.

Do you have an existing VM environment you are migrating or is this your companies first VMware implementation?

Carl

0 Kudos
MattG
Expert
Expert
Jump to solution

It depends on what the IO profile of the VMs on the LUN will be. If you have sVmotion then you could be more aggressive with your LUN size as it is easy to "fix" any sizing mistakes.

I used 300GB LUNs based on the fact that I want to keep my active vHDs per LUN under 20. If I add a VM that needs a large vHD, like a File server, then I give it a dedicated large LUN.

-MattG

If you find this information useful, please award points for "correct" or "helpful".

-MattG If you find this information useful, please award points for "correct" or "helpful".
0 Kudos
Rockapot
Expert
Expert
Jump to solution

Yeah i agree. I usually place between 10 and 15 VM's per VMFS volume/partition and anything with 1Tb or over on an RDM.

Carl

RParker
Immortal
Immortal
Jump to solution

These LUN sizes you see are from early adopters. Those LUN numbers are from older SAN devices, that probably were better with smaller LUNS. It's also based on disaster recovery.

If you have a LUN of 1TB and there are 50 VM's on it, then you lose 50VM's, initially, and age old thinking would have you believe that you have to restore the entire LUN first before you can use it, well that's not true. You can restore 2 or 3 VM's at once, and each one is done concurrently. When one is done, you can power it on right then, don't have to wait to restore the rest of the LUN.

The next problem with small LUNs is overhead, nobody thinks about that. If you allocate a bunch of smaller LUNs say 500GB, you have to establish a volume of say 520GB just for snapshots, and breathing room. So on a 2 TB volume, that's 100GB of wasted space, more or less. If however you make larger LUNs you conserve more overhead space.

There really isn't a problem with Large LUNs, it's comfort zone, and not taking into account advances in technology, for another thing. Early years you had drives of 8 and 16GB, those were the norm. 25 of those drives would yield 400GB, which coincidentally is the same size as what MOST people recommend, amazing.. They are just using what tradition SAYS they should.

I contend this is no longer the case, 300Gb drives are NOW the norm, so you are telling me it's STILL functional to use a 400GB LUN across 2 drives. I think NOT!.

Look beyond the tradition, open your mind. 25 drives is NOW 7.5TB. What's changed? NOTHING, just the drive size, so why NOW all of a sudden are the LUNs sized different? Well they aren't but peoples minds have not changed. That's what you should consider.

Make the LUNS doable. Make them easy for you to manage, but certainly there is nothing at all wrong using the same conventional thinking about number of disks, which is really what it's based on, NOT size.

Rockapot
Expert
Expert
Jump to solution

RParker,

Yeah true that you need to create LUN's depending on your environment and requirements however I dont beleive you are correct in that people are all selecting typical LUN sizes of say 500Gb "because tradition says they should", maybe some.

I take in to account individual clients needs and considering I have been consulting privately for SME's to large enterprise clients and am now consulting for European leader in virtualisation I can confidentely say that most clients adoption of VMware with LUN sizes between 500Gb to 1Tb is definately working.

So in effect the LUN sizes you are hearing people mention between the range of 500Gb to 1Tb are not because people are early adopters but because most clients environments fit well inside these LUN sizes.

I do agree with you in that there is nothing wrong with large LUN sizes however you also need to take in to account best practices from the various storage vendors depending on the clients virtualisation requirements.

Its not all about VMware..

As for 300Gb drive being the norm. Maybe where you are but not from what I am seeing. Yes I am not surprised when i see 300Gb volumes but no I am not seeing them as much as "normal" everyday VM's

Anyway..., as you stated "Make the LUNS doable. Make them easy for you to manage", I agree with this.

Not an argument, just my 2 cents which i feel is valid considering the numerous environments I have seen in the real world.

Carl

0 Kudos
RParker
Immortal
Immortal
Jump to solution

I have been consulting privately for SME's to large enterprise clients and am now consulting for European leader in virtualisation I can confidentely say that most clients adoption of VMware with LUN sizes between 500Gb to 1Tb is definately working.

yes, working based upon what measure? I realize it fits within their scope, but where does this range come from? Did you do the research, or are those numbers based upon what 'other' companies are doing. I don't think it was originated from them, therefore they had to hear it from someone else.. If you did the research that's different and you as a qualified tech knows what they should use, based upon need, but I see a lot of posts that discard why we make these volumes, it's disk number not size that ultimately we are looking for, performance > size. Too much importance on size as a matter of principle.

I do agree with you in that there is nothing wrong with large LUN sizes however you also need to take in to account best practices from the various storage vendors depending on the clients virtualisation requirements.

Also questions I have asked from other vendors, Netapp et. al have said that they have had 'YEARS' of research.. ok, so right there, their 'research' means traditionally speaking, which means they don't change what works.. which means they aren't thinking outside the box. Case in point. They based original LUN sizes on what?

Disk spindle count. So 25 or 20 disks comprises a Volume / LUN. Same diff, 20 36GB disks = 20 300GB disks. What's changed? Nothing. The disks are larger, so the LUNs must be larger as well. So since they don't change what works the ORGINAL formula early on was more disks = better performance. So the disk themselves have not changed, just the size.

As for 300Gb drive being the norm. Maybe where you are but not from what I am seeing. Yes I am not surprised when i see 300Gb volumes but no I am not seeing them as much as "normal" everyday VM's

OK, maybe norm wasn't the right word 'default' is what I am seeing from storage vendors, even local storage is based on 300GB, since SATA disks are well over 1TB as the default, some are even 1.2TB drives, unless you change the 'default' spec.

So if you have disks that big, it's only natural that the LUNS also will be larger to accommodate. Thats all I am saying. Smiley Happy

And look at it another way you have multiple volumes spanning those same RAID groups, right? So you put 2 or 3 or 5 or 6 LUN / Volume on those same RAID groups, so what is the difference between many LUN on a single RAID group vs 1 big LUN on the same RAID group, you are trying to do the same thing, you are using the same Disks, so you either build many LUNs across those same disks, or 1, the difference is management, but the risk is the same no matter how you look at it. The disks are utlized by 1 or more than 1 LUN in any case. The risk in this case is RAID group failure. You lose the contents of that RAID group should that RAID fail. So whether that RAID group was a set of LUNs or a single LUN the contents are what's lost, it won't matter how big those LUNs were, they are ALL gone. Just because they are split up doesn't make the risk any less. Still same data across the same set of Disks.

There are only so many shelves and so many disks you can use, at some point utilizing larger disks of the same quantity yet keeping the same size LUNs means space is wasted somewhere. That's what I am driving at...

So if you don't scale the LUN's with the drive space available, are you not wasting space and potentially keeping your environment limited because you want keep the LUNs small?

If you buy smaller drives, that's fine. Not knocking the disk size, just saying that not everyone should base their LUN's SOLELY on size, that's dependent upon the RAID and disk sizes within that RAID, that will determine the LUN size. Performance of the SAN is what volumes are all about, sizing is a secondary consideration.

0 Kudos
Rockapot
Expert
Expert
Jump to solution

Ok I see where you are going. Good posting. Its true that there are a lot of posts that discard the subject why LUN sizing provisions are made and i guess customers do tend to lean on the various storage vendors "research".

"what is the difference between many LUN on a single RAID group vs 1 big LUN on the same RAID group", well recovery. Whilst some customers do lock down environments down, some clients regain control of administrative accounts when environments are left to them after consultancies leave site. When the clients own internal policies fail admins can cause disruption to LUN's while they browse around so it is easier to rather recover from a single smaller LUN than that of a large 1Tb+ LUN in some instances. Smaller LUN's minimise the impact of multiple VM outages as opposed to multiple VM's on larger LUN's.

But yeah I see what you are saying..

Carl

0 Kudos