The Storage Hypervisor – Part 1
One of the keys – I think – to the benefits of cloud computing is abstraction. By abstracting our workloads from the boundaries of physical hardware we find ourselves able to do things and manage our resources in ways we haven’t before. When using hardware virtualization – such as VMware vSphere – the hypervisor not only frees us from the limitations of physical hardware boundaries and compatibilities, but provides a new entry point for management, creating a new paradigm.
Things we used to do with installing agents on each OS – such as monitoring, backups and more – can now be orchestrated through a single hypervisor. Provisioning an OS on bare metal used to be cumbersome and often complicated and expensive tools would be needed to do this at scale. With virtualization, not only can we push out a pre-configured OS in minutes, but with the right tools we can do the same for multi-tiered applications complete with firewall rules. These benefits enable us to position the IT organization for Agility. What if we could do the same with a storage hypervisor?
The concept of a storage hypervisor is similar – abstract away the traditional boundaries and limitations of storage with a unifying platform that enables a new paradigm of efficiencies leading towards agility. In fact, Wikipedia defines storage hypervisor as follows:
The storage hypervisor, a centrally-managed supervisory software program, provides a comprehensive set of storage control and monitoring functions that operate as a transparent virtual layer across consolidated disk pools to improve their availability, speed and utilization.
Before we begin looking at some of the benefits a storage hypervisor could provide, let’s first look at some of the requirements.
UNIFIED STORAGE
This term gets tossed around as much as “cloud computing” or “open source” and as often is the case, it can mean wildly different things. Perhaps the best way to illustrate the concept on unified storage is with contrast. Consider a SAN solution which runs three different operating systems – two versions of storage code, plus a 3rd Linux OS for management. To support all this, the SAN solution requires 9U of rack space, compared to just 5U for a competing SAN that uses only a single – and unified – set of code.
The picture above helps us to visualize just how much cost and complexity – and rack space – is reduced when a truly unified solution is used. Only the rack space savings is visualized, but other benefits are implied and these efficiencies will be detailed in future posts.
NetApp’s ONTAP as the Storage Hypervisor
NetApp’s FAS product line (which excludes the Engenio acquisition) is unified around the ONTAP operating system. One platform, one common set of code, from which to abstract storage and provide some outstanding features that provide new levels of efficiency – and when effectively combined can provide a platform for Agililty.
This series will explore these benefits in detail, but here’s just one to start with. With most SANs if you want to migrate from their “mid-range SAN” to the “high end SAN” you have to go through a costly and time consuming data migration to the new environment. Within the NetApp FAS series, you can upgrade simply by essentially swapping out (or replacing) the controllers. The existing disks and data stay in place allowing you to leverage your existing disk/SAN investment without any complicated data migrations. That’s a pretty big benefit! Or should I say Agile?
Earlier this morning Vijay Swami ( @veverything ) and I were discussing a Morningstar article which suggested that another storage vendor was more resilient and “bullet proof” because they had “developed a broad portfolio of solutions, each with a specific target” as opposed to NetApp having a higher “concentration of risk” to potential market “disruption” due to having a single platform.
Let’s think about this carefully for a minute. Is a vendor more “bulletproof” to potential market disruption because their product mix is a patchwork of different solutions that are less than fully integrated? I’d suggest that what’s paramount here is a vendor’s ability to effectively service different market segments (high-end, mid-range, low-end, etc.). If a vendor can effectively serve these same markets with a unified platform instead of a more fragmented offering with different technologies for different markets, that makes the vendor more agile. They will have a lower cost structure and support and integration is simplified. Perhaps the customer benefits the most by having a common OS across all classes of storage – even DR and backups.
In this series I’ll try to take a look at the ways in which ONTAP provides unique benefits for storage efficiency in a virtualized environment, and then build on these features to show how they can be combined with VMware and others in a virtualized environment to empower organizations with Agility.









Does your deliberate removal of EMC logos and choice not to actually type the letters EMC have a larger significance? It seems a little forced since my guess is that most of your readers understand the comparison you are making here.
@jdooley_clt No larger meaning. My intent was to make the contrast vendor neutral because the concept applies to all vendors. Yes the image is from an EMC product, partly because it offered one of the stronger contrasts, but I erased the logo as it was never my intent to compare with just one vendor. The MorningStar article choose to compare NetApp with EMC. There I was talking about their characterization of NetApp’s more unified product mix more than any other vendor.
Interesting. It didn’t seem, especially with some of the specifics you included in the post (software descriptions, rack configurations, etc…) as well as the reference to the MorningStar article, that you were trying to target anyone other than EMC. My mistake, I’m glad I asked!
Can you tell me how many interfaces it takes to manage Netapp Cluster-Mode? What about having to choose between 7-mode or Cluster-Mode now? Otherwise what’s the upgrade path? What if I need high performance SAN? I now have 3 different code bases. True?
If you want to talk about the benefits of Netapp, talk about where they really shine (Replication, NFS, etc) The “Unified” message they now have is just marketing.
Hey growler, thanks for replying. I agree with you about where NetApp really shines — and I’ll get to those in future posts — but I would suggest that it is the ONTAP platform that enables these capabilities, hence this introductory post.
As for your other comments about multiple code bases, I’m not sure I follow. ONTAP 8.1.1 supports multi protocols, FlashPools, 7-mode, cluster mode, and across the entire FAS product line. Granted different servers might be configured differently based on their role (cluster, DR, Tier 1, etc.) but its still all behind a common and unified code base.
@BlueShiftBlog So would you call a platform that needs three different interfaces to manage “Unified”? So can we call EMC VNX “Unified” because it has Unified Management?
There is multiple ways to achieve unified and Netapp and EMC tackled this in different ways,.
You can put “unified infrastructure” in front of a single device or a Rube Goldberg machine with 30 moving parts. An extreme example but I think you get the point — one is more unified and there’s real benefits behind that. Some of what’s possible with a storage hypervisor like ONTAP is in the above post — but the best will come far later in this series
@friea Thanks!
DISCLAIMER: Nick from NetApp
So much I want to say here. First and foremost, excellent post, excellent comparison points. At the end of the day, taking everything out of the fridge and wrapping it up in Saran Wrap is not Unified. That will never, ever scale.
growler made some good points about our current product line. And while your short-term observations are close, you’re thinking about 7-mode and cluster-mode the wrong way. This is a transition, not an addition. The end of 7-Mode is imminent. There will always be those customers that keep it around and refuse change, but in the future, everything will be Cluster-Mode. 2-node Cluster-Mode arrays will replace the typical HA pair, enabling future growth and scale-out by simply adding nodes. Again, you’re right, there is no “magic bit” to flip to make one go from 7-to-Cluster, but it’s not as big of a deal as “the industry” seems to be making it out to be. Failover your 7-Mode HA pair, reconfig one controller to Cluster-Mode, migrate volumes, reconfig second controller and add to cluster. Done. Am I over-simplifying? Maybe a tad. But these types of comments do nothing good and only over-exaggerate the rumor mill. If you want to play with Cluster-Mode, and see how it truly works, we’ve got a vSIM for that.
DOT 8.1+ is not just about Cluster-Mode. Flash Pools, NFS VAAI, 64-bit aggregates with 100TB, Infinite Vol’s, the list goes on. Sure Cluster-Mode is a huge milestone almost 10 years in the making, but don’t be short-sighted, guys.
Lastly, when did this misconception come around that NetApp wasn’t capable of running high-performance SAN workloads? Not only can it do that, but we’ll also run NFS, CIFS/SMB, B&R, DR, etc, ALL ON THE SAME BOX! And using the same code/interfaces on any box from the upcoming VSA, all the way up to the big-boy FAS6280. Did I mention we can also virtualize non-NetApp arrays on that same box as well?
This is all stuff you guys know, so why does it always have to be a focus on the negative, and not a congratulations on the accomplishment? Sure, it’s got a ways to go before it’s perfect, but so did VMware with VI3. Think back to before we had vCenter, NFS, etc. There were flaws and growing pains before we got to that happy point with vSphere 4.1.
I’d rather focus my efforts on educating customers rather than coming onto well-written posts such as this one and dispelling FUD.
-Nick
Pingback: The Storage Hypervisor Part 2 -- Flash Pool | Blue Shift
Thanks for RTs! @Mandivs @Arjantim @TheCloudNetwork @that1guynick @friea @diversetips
EMC is the biggest pos company in the storage space.
It came with 20 different discs each containing a 2 KB file called a damn enabler and they forgot(wink wink) to put the main enabler to make any of this thing work. They short sell the whole thing and want to pay for the professional service for installation. Deliberately withholding information . Extremely devious marketing.
On the phone with their stupid suppport people and sales for a week. Still could not get the damn thing going. All they want is to pay them in 10 thousand$ blocks for their professional service people to come in.
We finally threw all of their stuff out for an HP solution and got it running in half a day from racking to serving storage to a VM cluster.