Compellent wins Microsoft Storage Partner of the Year

I wasn't commenting that NetApp has a similar ILM product, that's why I said it's Compellent special sauce.  While you are correct about space guaranty on the volume, most NetApp admins I know only guarrenty on business critical apps like SQL and Exchange....NAS/NFS/CIFS is usually only file or none....while its not "purely" thin-provisioning, its still possible.  In reality, does anyone really want to over-commit storage, how valuable is thin-provisioning in the end.  I will tell you this, thin provisioning is automatic with NFS on VMWare, slap ASIS on top of it and I'll take the 40-60% reduction in storage over ILM any day of the week.

 See you next week!!!

I've got a few comments for you guys.  First off, if Compellent didn't have a command line I wouldn't have bought it.  The actual command to create a new volume on Compellent looks like this: "volume create -name CommandLineRox -size 2000g -readcache true -writecache true".  You can add parameters to tell it which server to attach the LUN to and configure the block level ILM all in one command.

On Thin Provisioning, Compellent still has the TP advantage for VMWare that blows away the benefit of using NFS.  The main reason ESX guys are starting to use NFS is the enormous waste inside a VMDK file.  With Compellent's TP, I can create any size VMDK and I don't consume physical space on the SAN LUN until I write data inside the VMDK.  Compellent can recognize white space (sequential zeros) inside a VMDK file and ignore it.  They do the same thing with white space in a SQL or Oracle DB.  So after I create my 100g VMDK file on a Compellent 4Gbps FC attached LUN, I've consumed 0g of disk space.

Compellent is the only system that has this kind of block level intelligence and an uber command line on top of loads of special sauce!  TP and ILM just scratch the surface.

By the way, for those of you who think that NFS is faster than an FC LUN, VMWare disagrees with you.  Take a look at "Comparison of storage performance protocols in ESX Server 3.5", http://www.vmware.com/files/pdf/storage_protocol_perf.pdf

I would never use NFS for my production VM's.  Of course, on NetApp, NFS is definitely faster than the SAN emulation they use to mount an FC LUN.  NetApp is just a NAS after all.  So I guess it depends on which SAN you use.  If you use NetApp, I'd recommend NFS for VMWare.  If you use a native FC SAN, then FC will blow away NFS performance every time.

-Jason

Jason,

 I evaluated Compellent about 6 months ago and was told by the SE and Sales SE there is no command line interface, if they changed that, great!  However when I looked at it, is wasn't available.

As for the VMWare doc...thats a completely loaded document because they only cover a single ESX host.  The original design of NFS was having multiple hosts reading and writing from the same target.  Once you go beyond 5 ESX hosts accessing the same NFS mount, performance beings to equal that of 2GB FC - provided you designed your ESX infrastructure correctly.  By using NFS on multiple ESX hosts, you elimate the SCSI reservations used by VMWare on FC with VMFS.  Using NFS on NetApp gives you the same exact thin-provision you described with Compellent.

Lastly, I can only comment on NetApp, I haven't seen NFS on other storage platforms, but another big reason the NetApp NFS gives you FC like performance is their WAFL filesystem.  There are other advantages as well, like being able to snapshot all the VMDK's on the storage side without need to do anything on the ESX side, again, you're eliminating VMFS.  Or being able to share out your NFS mounts via CIFS and mount your snapshots on media servers for backups, and my favorite ASIS (automatic single instance storage) that removed redundant blocks and replaces it with metadata...I've seen almost 50% reduction in storage on VDI with ASIS.

Just to be clear...I'm a fan of Compellent, I think they have a nice product...but storage vendors come and go except for the big players I mentioned earlier.  It won't be a shock if one of those bigger players buys Compellent just for their ILM technology.

Hey Michael,

See ing as you ahve some experience with Compellent,. Are you familiar with Equallogic? We have  afew of their arrays here and really like the simple management, and addons in thier Host integration toolkit. They will also do some ILM and automatically move things around to different speed disks as needed (although we have yet to use this as all our disk is the same speed).

 Since Dell grabbed them up we are now a little scared to buy any more.

Have any comparisons betweene them that you feel like sharing?

Right on dude....EQ uses Copy on Write for their snaps which pre-allocates like Mike said...Compellent uses Allocate on Write which only right writes blocks when they change.

 The only problems I've "heard" - I haven't seen this in action, with EQ is keeping a large number of snaps starts to impact performace of the production LUN.