Storage virtualization doesn’t exist

Posted by Alessandro Perilli   |   Monday, July 21, 2008   |   15 Comments

virtualization.info was launched in September 2003 and in almost five years of activity we received thousands of emails and comments.
Some of them carry on recurring questions so that at this point we can safely compile a top 3 FAQ:

  • Which processor grants the best performance for virtualization between AMD and Intel?
  • Which storage array is the best for virtualization?
  • Why doesn’t virtualization.info cover storage virtualization?

Let’s save the first two for better times (we are actively working to provide an unbiased answer to those questions).
It’s time to answer the third one: virtualization.info doesn’t cover the so called storage virtualization because at today this term doesn’t mean anything.

Unlike what happens for hardware virtualization, OS virtualization and application virtualization, the storage vendors seems unable to find an agreement on the definition.
The term is abused in almost every press announcement and it can refer to at least ten different approaches.

Depending on the vendor you are talking to, storage virtualization is the abstraction of the directories, of local volumes, of the remote volumes, of the array, etc.
So even well known concepts like RAID or distributed file systems become storage virtualization and get sold as brand new, cutting-edge technology enhancements.

For years, starting in early 2003, the worldwide press followed the marketing folks claiming the advent of storage virtualization as the new mainstream. But the reality is that the market is so incredibly confused that potential customers just distrust the technology.
How can we have confidence in something that not even the vendors recognize in a univocal way?

It’s evident that the vendors are using the term in a wild way hoping to raise the same interest that server virtualization created in the last few years.
What they are missing is that they are obtaining exactly the opposite effect.

Server virtualization (and VDI) are the best thing ever happened to the storage market in years.
Thanks to VMware, Citrix, Microsoft and all the other companies we track on virtualization.info the need for shared storage reached record levels so that even the SMBs have to contemplate it now.

There’s no need to use the term storage virtualization and confuse the market to sell more.

We are so glad that the independent analysis firm Burton Group has a similar opinion:

So what's important? Forget about the term 'storage virtualization'. Focus on function and features. Determine what you are looking for in the ultimate storage ecosystem. Do you want to provide nuts to bolts information lifecycle management? Ok then, look for storage vendors who can move data auto-magically from storage tier to tier with some content management thrown in. Or are you tired of managing volume capacities? Then get yourself into thin provisioning. Need volumes to follow around server virtual machines? Look for storage that can flit along with the VM. Need to backup and archive?  Get data deduplication. Consolidate all those disk arrays you've got into a common management point with a few extra features thrown in? Yes we can. Whatever - you get the idea. Call it what it is, describe the function you're looking for and find vendors who can deliver. If the storage vendor wants to call it virtualization, then great, give him a knowing smile, pat him the back, but ask for a firm bid…

At virtualization.info we are ready to talk about storage virtualization, as soon as the industry stops to mislead the customers.

15 Comments

Blogger Claiton Z Monday, July 21, 2008 10:48:00 PM  
Do you know about IBM SAN Volume Controller and over provisioning ?
Anonymous Anonymous Tuesday, July 22, 2008 3:48:00 AM  
Truly excellent article. It's about time someone called the industry on this terminology abuse. Taking any word and placing it before the word "virtualization" in hopes of luring/confusing customers is misleading at best.

I'm sick of reading how RAID controllers, fibre channels interfaces and/or NFS is storage virtualization. That's all a load of bollocks.

Great analysis!
Anonymous Val Bercovici Tuesday, July 22, 2008 6:58:00 AM  
I couldn't agree more. See

http://blogs.netapp.com/exposed/2008/07/virtualization.html

Even though NetApp participates in the storage virtualization market via our cool V-Series technology, it's abundantly clear to me that the server virtualization market set the bar high for proper usage of the term "virtualization".

Perhaps we in the storage industry should better articulate the Value our "V-something" solutions deliver. For example, adding first-class dedupe, thin provisioning, thin replication, near-cdp via NetApp snapshots, WORM, D2D backup and an optimized NFS option for VMware & Citrix is what NetApp offers.
Anonymous Anonymous Tuesday, July 22, 2008 8:43:00 AM  
Do you know HDS USP-V and storage provisionning ?
Anonymous Anonymous Tuesday, July 22, 2008 1:02:00 PM  
I agree, take a look at IBM SVC and DataCore
Anonymous Anonymous Tuesday, July 22, 2008 3:54:00 PM  
I agree with the overuse of acronyms and cool souding marketing terms. I also agree with concentrating on function BUT, we also must define what virtualization really means.

For different technologies, virtualization could mean different things. For servers, virtualization means, hardware independence. For applications, virtualization means, OS and location independence. Similarly, for storage, in my humble opinion, virtualization means, independence from the type of devices.

So what if I have multi tier storage in my data center? If I can connect it to my servers (virtual or physical), present it to the OS (again, virtual or physical) and applications (virtual or physical) can store their data (must be physical) on these devices (regardless of location, type etc) and then able t read it back, I would consider this "function" storage virtualization. So perhaps we must look for storage virtualization outside of hardware vendors?
Anonymous Anonymous Tuesday, July 22, 2008 5:41:00 PM  
Perfect point.
I was telling to our marketing that storage virtualization is a hype.. they did not believe me
Anonymous Alex Barrett Tuesday, July 22, 2008 8:23:00 PM  
Alessandro,

I disagree on your point that "storage virtualization" doesn't exist, but no matter. It's just funny to me how things have changed. In 2002, I used to cover storage, and there was no problem storage virtualization couldn't solve. Then as quickly as the storage virtualization hype took off, it plummeted, to the point where storage vendors treated "virtualization" as a dirty word. Check out this 2003 article by storage analyst Arun Taneja, where he refers to it as "the V-word."

Now, storage vendors are falling all over themselves to brand their technology "storage virtualization." Luckily, IT people are pretty smart.
Anonymous Olivier Withoff Tuesday, July 22, 2008 9:41:00 PM  
The Wikipedia article on SANs, seems to think the term means something:
"Storage virtualization refers to the process of completely abstracting logical storage from physical storage. The physical storage resources are aggregated into storage pools, from which the logical storage is created. It presents to the user a logical space for data storage and transparently handles the process of mapping it to the actual physical location. This is currently implemented inside each modern disk array, using vendor's proprietary solution. However, the goal is to virtualize multiple disk arrays, made by different vendors, scattered over the network, into a single monolithic storage device, which can be managed uniformly."
http://en.wikipedia.org/wiki/Storage_area_network#Storage_virtualization_and_SANs
Anonymous Anonymous Wednesday, July 23, 2008 12:26:00 AM  
actually on cpu, AMD already sell their cpus with NPT (nested page table), compare to intel where EPT is arriving end of 2008. i think that AMD virtualisation is actually a bit better for the moment (and they have full real mode emulation too !). intel is late 2nd on the cpu virtualisation feature race.
Anonymous Chad Sakac Wednesday, July 23, 2008 1:57:00 AM  
Alessandro, like you and I discussed face to face - I (and most of my EMC compatriots) agree.

But I will speak for myself....

**Storage Virtualization doesn't exist**!

ah... feels good :-) Well, it's true at least how vendors (ouch, that includes us - but prime marketeers on this front are IBM SVC and HDS - just look at the comments :-) shlep the goods. Storage Virtualization is positioned as a panacea, and something intrinsically linked to server and application virtualization, and implying the same use cases and benefits (and that you're a tool if you're not doing it).

I do agree with Oliver, and agree with the Wikipedia definition.

Storage abstraction (all the levels described except ACROSS platforms) have existing in all sorts of arrays for eons, and that's goodness.

ACROSS platforms is the only (relatively) new thing - but the primary consistently beneficial use cases we've seen are massive data migrations.

There are some customers literally so large they perpetually migrating data. For those customers (a relatively narrow band) the incremental value is immense, since a device that decouples the storage presentation from the array means that the last setup of the migration (switch host view from one LUNID to another) is eliminated. Otherwise they are perpetually managing scheduled downtime for migrations, and this new class of capabilities eliminates that case.

For almost everyone else, they get what they need without anything fancy - just using any serious array on the market (everyone thinks they do that bit better than everyone else)

Consistent management can be gotten via management tools that support heterogeneous configs (EMC's got 'em, NetApp has got one, Symantec has some). My two cents worth.

While Chuck and I disagree as much as we agree, when we agree (and disagree), it's vehement :-)

Here, I agree with him (and you can see that we're saying this too!):

http://chucksblog.typepad.com/chucks_blog/2008/06/post.html
Anonymous Gene Ruth, Burton Group Wednesday, July 23, 2008 8:23:00 PM  
As the writer of the Burton group blog referenced here, some comments:
“storage virtualization doesn’t exist” is a little over the top, it certainly does exist. The trouble is it means many things to many people – storage virtualization comes in so many forms that the term is useless. If one takes a functional viewpoint, then all the following can legitimately be called storage virtualization:
• Creation of storage volumes (including file systems) based on physical storage devices such as a disk RAID array
• Aggregation of storage volumes (LUNS, filesystems) within or across storage subsystems
• Automation of storage volume creation, placement and management, e.g, LUN creation, thin provisioning, data de-duplication and tiering.
• Aggregation of disparate vendor equipment into a unified storage resource presentation
• Encapsulation or translation of storage connection technologies, e.g., iSCSI to FC or FC to SAS translation.
• Impersonation of storage devices such as a virtual tape library with a disk array
• Creation of file systems, both singular and distributed, such as offered by network attached storage via NFS or CIFS

That’s quite a variety of functionality to simply label as “storage virtualization”. So therefore my original blog comments. Once users identify their motivation such as:
• the simplification of storage management,
• improved utilization,
• extended function of commodity storage,
• isolation of storage service from the hardware,
• or isolation for security reasons,
Users should identify the functionality they really need and search out products to meet the need.
As the writer of the Burton group blog referenced here, some comments:
“storage virtualization doesn’t exist” is a little over the top, it certainly does exist. The trouble is it means many things to many people – storage virtualization comes in so many forms that the term is useless. If one takes a functional viewpoint, then all the following can legitimately be called storage virtualization:
• Creation of storage volumes (including file systems) based on physical storage devices such as a disk RAID array
• Aggregation of storage volumes (LUNS, filesystems) within or across storage subsystems
• Automation of storage volume creation, placement and management, e.g, LUN creation, thin provisioning, data de-duplication and tiering.
• Aggregation of disparate vendor equipment into a unified storage resource presentation
• Encapsulation or translation of storage connection technologies, e.g., iSCSI to FC or FC to SAS translation.
• Impersonation of storage devices such as a virtual tape library with a disk array
• Creation of file systems, both singular and distributed, such as offered by network attached storage via NFS or CIFS

That’s quite a variety of functionality to simply label as “storage virtualization”. So therefore my original blog comments. Once users identify their motivation such as:
• the simplification of storage management,
• improved utilization,
• extended function of commodity storage,
• isolation of storage service from the hardware,
• or isolation for security reasons,
Users should identify the functionality they really need and search out products to meet the need.
http://dcsblog.burtongroup.com/
DataCore is storage virtualization. They pioneered the entire concept of storage virtualization over ten years ago. I believe they are qualified to define what storage virtualization is.
Blogger Kashif Shaikh Thursday, July 24, 2008 4:52:00 PM  
Virtualization - everything is virtualized to some degree.

Your programs running on your computer are running in a virtual address space. The CPU is designed to provide you with the illusion of a single address space using page tables and a memory management module.

You hard-drive is virtualized to look like a flat file. In reality hard-drives convert LBA addresses to some address the hard-drive can understand (position head on platter 2, read the sector).

Your network is virtualized as a socket connection with two end points providing reliable transfer, regardless of the network card type (ethernet, fibre channel, local loopback, etc).

Your storage device is virtualized to look like a reliable hard-drive with the ability to read/write multiple hard-drives without the OS knowing about those drives.

The reason storage is touted so much as virtualization is because 99% the SCSI protocol is composed of READ AND WRITE messages and is a request/response protocol. Not hard to virtualize.

You can implement the SCSI protocol to read files from Mars Rover and storage vendors would still tout it as "amazing virtualization"!
Anonymous Massimo Re Ferre' Friday, July 25, 2008 11:50:00 AM  
Alessandro,

I have to disagree with this one this time.

If you take the two realms that are Servers and Storage I don't see how they are different in the way vendors mislead the market.

Like there are multiple meanings of what virtualization is in the storage realm... so there are multiple meanings of what virtualization is in the Servers realm.

The only difference is that for Servers someone sat down and started classifying everything (i.e. hw virtualization, os virtualization, application virtualization etc etc) while for Storage no one has started to do that (in an agnostic manner at least). I think you have an opportunity here to lead that effort based on the comments I read.

Yes Raid is a form of virtualization and I agree that it confuses because it has been around for decades .... but hasn't memory too been virtualized on servers for a long time? How long has the pagefile.sys been around on Windows? Isn't that a form of virtualization on servers where you take a given resource and overcommit/virtualize it?

Massimo.

Add New Comment