Introducing The VNX Family

Among all the many storage announcements from EMC today, perhaps the most impactful to many is the new VNX family.  

Certainly, it’s been the subject of the most speculation.

The new VNX models can best be seen as a linear extension and convergence of the wildly successful CLARiiON and Celerra lines – but goes farther than either one individually.

And the new VNXe is in many regards an entirely new product designed to serve an entirely new market segment.  

Both products will likely be extremely hot topics in the industry storage discussion during 2011 and beyond.  

Fasten your seatbelts :-)

The Storage Array Landscape

Perhaps the most hotly contested portion of the traditional storage market is the dual-controller mid-tier market -- it's the bread and butter of external shared storage.

Dual controller designs provide essential redundancy that single controllers can't, albeit at additional cost.  That being said, any dual-controller design that experiences a single controller failure operates at 50% (or less) of its rated performance -- occasionally making a case for more sophisticated designs that use multiple controllers such as EMC's VMAX.

Storage vendors in this particular segment have to bring their best game, since competition is strong -- and occasionally very loud! The VNX Series announcement represents an entirely new generation of EMC's core products in this important category, and thus becomes a new benchmark for comparisons.

A Bit Of Context

This slide from the customer deck pretty much sums up the frequent challenges in this important category.

I like the term "affordable" because it captures the reality of how people often look at the acquisition cycle in this segment.  

Infrastructure funding is usually limited to a per-project basis -- even though an expanded view of the landscape and timeframes might lead to other alternatives. 

The reality is that there's only so much money in the budget for the next storage acquisition -- and that's that.  Storage platforms with initial acquisition costs outside of the allocated funds just don't make the cut -- regardless of their potential strong points.

I'd argue that the concept of "simplicity' is currently being redefined in the storage marketplace.  It’s always been important – but I think now it’s going to start to be measured differently.

The new measurement?  It's going to be how much can you get done -- and how quickly -- without any specialized knowledge, training, documentation, support, etc.  Just to make the point -- there's no training, manuals, certification, etc. for consumer products like the iPad :-)

EMC believes that storage in this category should ideally be easily usable by an IT generalist, vs. trained storage experts.  

This means workflows supported in the context of the application or task, avoiding the use of storage jargon, intelligent defaults throughout, and so on.  

The model is much like setting up a big-screen TV -- you want to get to the good part (using it!) rather than wandering around in multiple menus and manuals.  

That being said, there still needs to be deep menus and functions available if you really want to go there and satisfy your inner geek :-)

"Efficiency" is an ever-popular topic in the storage world.  Over the last few years, so many technological advances have conspired to make "effective usable capacity" the preferred metric, rather than simply considering raw capacity.  This technology list includes flash, compression, deduplication, thin provisioning and so forth.

Cost-per-usable-terabyte has dropped much faster in the last year or so than the raw cost of media.  My advice?  Consider asking your vendors to quote in "effective usable capacity" for your application environment vs. simple raw capacities.

"Power" is a term worth exploring.  More than mere performance, I tend to think in terms of ability to add more capacity, more features, more functionality -- as well as performance -- while keeping your initial investment intact.

From my perspective, the new generation of Intel processors from Intel can provide almost absurd levels of storage controller performance.   There's far more oomph there than most people could ever use on Day 1.

But the workloads people tend to throw at storage arrays tend to grow substantially over time.  The workload growth is usually the result of no one single thing: just an unending  river of new applications, more users, more data, more functions, etc.

The new VNX series has plenty of power in reserve for these all-too-common scenarios.  This usually translates into a longer useful life for the asset vs. being forced to upgrade the controller in a shorter timeframe.

The New VNX Family

Here they are -- the new VNX series.

The VNX effectively replaces both the CLARiiON and Celerra lines going forward.  Users of each  will find strong commonality with the environments they're used to -- plus a substantial number of new features and capabilities.

The new VNXe is a story in itself -- a ground-up new platform, built with proven EMC code assets, and targeted at fast-growing segment of the market with unique requirements.

The CLARiiON, Celerra, VNX and VNXe all share a common management environment -- Unisphere, successfully introduced in mid-2010.  

Customers who have already made the leap to Unisphere (and have most likely come to appreciate its unique strong points) will find the user experience very familiar on the new platforms.

More on speeds and feeds in a moment.

Storage Is (Simplified) Software

As storage has become progressively more sophisticated over the past few years, the natural tendency was to simply create new storage software packages and associated titles that reflected all the new functionality.

Over time, though, this wasn't ideal.  

Too many titles, licenses, etc. created unneeded complexity and costs for our customers, our partners and ultimately ourselves.

One of the brighter parts of the VNX announcement is the complete repackaging of software functionality into a very small number of "software packs" that have all the related functionality that just about everyone could want or need.

We think that making storage simple involves much more than a great user interface: we're systematically looking at all aspects of how storage is specified, acquired, deployed, operated, upgraded, supported, and so on.

There shouldn't be a need to pore through a long list of individual software titles and descriptions -- you should be able to get what you need: simply and efficiently.

I'd predict that -- before long -- this sort of uber-simplified storage software packaging becomes the norm in the industry.  

Frankly, it's about time.

Considering The New VNXe

Although the brand-new VNXe can justifiably be called part of the VNX family, it's different -- and interesting enough -- that it's worthy of some extended discussion.  

In addition to being a killer product in a market segment that's relatively new to EMC, it expressed several design concepts that will inevitably find their way to other storage platforms over time.

The VNXe is laser-focused on an audience we've come to call the "IT generalist".  

These are the hard-working individuals who are responsible for keeping a wide swath of IT infrastructure up and running, frequently in a smaller setting. 

And a whole lot of IT gets delivered by these folks.

There is no time -- or interest -- in learning all the details or specifics of any one technology: storage, servers, networks, etc.  

These people have to get their job done quickly and efficiently, and without the need for specialized training, certifications, detailed manuals, etc.

As we learned more about these specific customers -- and their unique requirements -- there emerged a frequent scenario.  

The IT generalist was getting along fine with internal DAS storage (storage embedded in individual servers), and - one day -  there was some sort of tipping point that forced consideration of external storage.

Maybe it was growth and/or performance requirements.  Maybe it was the need for a clustered approach for either more performance, load balancing or high availability.  

Or maybe it was the realization that storage bound into specific servers was getting hard to manage, and even harder to pool and share.

Regardless, these went looking for their first shared storage device.

The attributes of "simple, efficient, affordable" take on a special and focused meaning in this scenario.

The picture of the Apple product isn't accidental, nor is it pandering.  The EMC design team frankly took great inspiration from Apple's approach to simplicity, integration and elegance.  

I'll leave it to actual users to grade us as to whether we've approached that lofty goal or not.

Breaking The Price Barrier

During the primary research for the VNXe, it became clear that most of our targeted IT generalists were familiar with EMC, thought we made great products, had excellent support, etc. 

They just thought they couldn't afford it.

Well, with the VNXe, usable configurations start at well under $10k.  

Not only is the product completely re-engineered to be more cost-effective using the latest technology, the entire ecosystem has been re-engineered as well.

It's now far more cost-effective for our partners to get trained on the VNXe, far easier to configure and propose, amazingly easy to install and get productive, simpler and more cost-effective to service and support.

Indeed, I don’t think it will it be too long before we see the economic model of the VNXe find its way into other storage market segments.

A Clean Sheet Of Paper?

You might be tempted to think that the VNXe is nothing more than a simple repackaging of existing EMC products and technology.  

You'd be somewhat right, and somewhat wrong.

The hardware is a brand-spanking-new design optimized around ethernet on the front end, and 6Gbps SAS on the back.  And the software is a new platform that enables the straightforward re-factoring of EMC's vast library of storage assets.

Indeed, the story behind the development of the VNXe is worthy of an extended discussion -- I'll be posting a series of blog posts on the story-behind-the-story this before too long.

Making The Array Smart

You really need to spend some time with the VNXe to fully appreciate just how smart and simple the user experience has now become.

One stand-out example is instant access to configuration expertise through a series of configuration wizards.  

Each and every one reflects EMC's proven best practices in storage configuration -- although you wouldn't know that to look at them.

For example, setting up a few hundred Exchange users involves a short and simple dialog around how many users, how big the mailboxes, what version of Exchange you're using -- and that's about it.

All the gory stuff -- RAID protection, thin provisioning, striping, etc. -- all handled transparently.  Of course, if you're so motivated, you're free to adjust specific settings and defaults.

But -- here's the beauty -- you don't have to really know anything about storage to set up a picture perfect Exchange (or VMware, or Microsoft Hyper-V, or file sharing etc.) environment.

The same experience continues on to the main user interface: you get a "big picture" that gives you a quick visual status (capacity, alerts, etc.) with all the typical functions logically presented -- with everything just a click away.  

What I really appreciate is that all the basic reporting functions are essentially built-in: there's nothing really extra you have to do to get simple yet useful reports on how the capacity is being used, how much spare capacity remains, and so on.  

A New Take On Support

We, as vendors, tend to make support reasonably difficult for IT specialists.  As we thought about meeting the needs of IT generalists, we quickly realized that the support model had to change -- and drastically.

First, everything you're likely to need has to be accessible from one place -- and one place only: training videos, how-to notes, community access, documentation, downloads -- even access to real, live people if needed.

Second, there's a progressive layering of support detail as you work through an issue.  Simple solutions to common issues are presented first.  More detail is shown as the user drills deeper into a problem.  There's a real concern evident about overwhelming people with too much irrelevant detail and jargon.

Third, there's a nice visual interface to the hardware aspects of the array.  You get nice, detailed renderings that show controller configuration, connectivity, power supplies, fans, etc.  When there's a hardware-oriented alert, it takes you to the exact place on the device where there's a potential problem.

If you need a quick refresher on how to, say, change out a power supply, there's an on-demand video that will tell you precisely what you need to know -- in addition to step-by-step on-screen visual instructions that can walk you through it.

VNXe: Speeds And Feeds

OK, the hardware types probably will be interested in this chart -- and there are a few things to note.

Perhaps most controversial will be the absence of traditional FC support at the outset.  This is a market segment that overwhelmingly prefers ethernet – and the cost-effective 1Gb kind at that.  

Although the design permits adding additional flavors of front-end I/O modules over time, the team came to the conclusion that -- yes -- 1Gb ethernet was all that was really needed here at the outset.

Besides, making that key decision has kept cost and complexity down -- not only for us, but for our customers and partners as well.

Note the potential for a seriously large non-volatile storage cache (24GB) on the larger VNXe 3300.  As well as the 6 Gb SAS back-end.  Not to mention the healthy drive counts (96 and 120) respectively.  This is no toy storage array.

I am told that integrated support for enterprise-class SSDs (as found on the VNX) is being worked on as a software enhancement -- I'll share more as I hear it.

Likely Impact Of The VNXe

There's every indication that the new VNXe will be a big winner in this important market segment.  

Our target audience told us very clearly – they didn’t feel their needs were being well-served by the options that were already out there.

When we put the VNXe side-by-side with the usual competitive suspects, the comparisons are always the same: the VNXe is more powerful, more affordable, more functional -- and, above all else -- dramatically simpler to buy, install, operate and support.

Simple -- and solid, thanks to its EMC heritage.

My guess is that, before long, it becomes the new standard that others will aspire to.

And On To The VNX

Many of the same themes found in the VNXe are also found in the VNX, but -- here -- we're usually talking more sophisticated and complex requirements.  

As mentioned before, the VNX replaces both the successful CLARiiON and Celerra lines with a single, unified platform that is essentially a superset of both.

The important -- thought subtle -- theme here is the continued virtualization of storage: the ability to create pools of capacity that automatically respond to changing application demands.

Although it might sound complex to some, in practice this is proving to be a far simpler (and efficient) method of managing storage vs. the traditional application-at-a-time hand-carved approach.

A storage pool is intended to deliver a certain service level (say "Silver") for a given set of applications (say "SQLserver").  

The pool itself is a blend of different storage types: various mixes of SSD devices for performance, SATA and/or SAS drives for bulk capacity, and FC drives sort of in between somewhere.

Through the magic of software (FAST VP in this case), each pool automatically re-shuffles active and inactive data between media types to achieve both service level and cost objectives without a lot of monitoring, tuning, adjustments, etc.

Far less expertise is required to set up and monitor these environments as a result.  

Surges in demand are either handled automatically using resources in the pool, adding more resources (e.g. SSD) to an existing pool, or non-disruptively moving the application to a higher grade pool, e.g. "Gold".  

Generally speaking, there's no need to over-provision performance and/or capacity (and associated costs!) just to play it 'safe".  

And that’s a big win in many environments.

And Performance Is A Big Story Here

The VNX is a new hardware design that is built around recent Intel Xeon 5600 multi-core processors.  Add in the new 6 Gbps SAS back-end, intelligent use of SSD technology, and we're talking eye-popping new levels of performance.

The "3X" claims you see throughout (as compared with the previous generation from, say, a year ago) are widely applicable, and even understand performance boosts in some cases.

The other cool magic here is EMC's "flash first" approach in the VNX.  SSD devices can either be used as non-volatile storage cache, or as a high-speed storage device, or any combination of the two.  

Many workloads respond amazingly well to hundreds (or thousands!) of gigabytes of flash-based storage cache that can be used for both reads and writes.  Others respond well to having their "hot data" automatically placed on SSD storage devices.

Just about every enterprise workload on the planet responds to some combination of both -- and that's what the "flash first" approach is all about.  

There's no need for pedantic arguments around the theoretical superiority of one approach or another -- users are free to use any combination of caching and/or SSD target approaches that do the job for them -- all managed automatically.

This new "performance dividend" shows up just about everywhere: database environments, virtualized workloads, email, web servers, etc. etc.  

If your application does mixed I/O (and that's largely the case in enterprise settings), it'll take a while to get used to these new levels of application performance that result.

The bad news? 

We're going to keep our EMC performance testing teams very busy for the foreseeable future precisely re-characterizing the significant and meaningful new levels of application performance for the hundreds and thousands of workloads that are out there.  

And IT customers had better be prepared to do some quick explaining as to why everything is now flying along so much faster than before.

But I think just about everyone will be pleased :-)

Simplicity Matters Here As Well

Many of the simplicity themes that are in evidence on the VNXe are also evident on the VNX as well.  

In particular, the "Unisphere experience" is mostly similar, although aimed at more sophisticated use cases and requirements.

The same Unisphere environment here has been extended to reflect the realities of larger settings: more arrays (old and new!), replication engines, etc.   In particular, the integration with RecoverPoint's CDP/CRR engines is getting a lot of attention.  

RecoverPoint was always popular for both its power and simplicity -- with this announcement, you can clearly see it becoming more fully integrated with the VNX array.

Although there's a decent amount of "one dashboard" integration with the current versions of Unisphere, this will undoubtedly be an area of rapid progress as EMC brings more and more into the Unisphere storage management platform.

As one example, total amount of effort for frequent tasks (e.g. setting up a high-performance SQLserver environment in this example) are substantially and noticeably reduced.

And, Yes, There's Cool VMware Integration

EMC has been the #1 choice for virtualized environments for quite a while now, and -- with the new VNX -- the VMware integration goodness just keeps coming in big chunks.

I'm sure Chad and others will delve into specific details, but -- at a high level -- the VNX not only boasts support for all the cool VAAI accelerators, but also supports a VMware-centric model where the VMware administrator can manage most aspects of storage without the need to get the storage administrator in the loop.

The storage administrator creates pools for use by the VMware team; who then uses EMC's Virtual Storage Integrator that appears as part of the integrated vCenter management environment.

No need for extended back-and-forth between the storage team and the VMware team -- everyone has their own streamlined process and workflow.

And On To The VNX Hardware

Although the family starts at the 5100 model, it gets more interesting at the VNX 5300 model and grows from there.

Drive capacity range from a max of 125 in the VNX 5300 to an extremely robust healthy 1000 drives in the VNX 7500.  

Storage processors (always 2!) range from 4 to 10 I/O slots as you move up the family.  Block protocols can be a combination of FC, FCoE or iSCSI.  DRAM-based storage cache ranges from 16GB to a healthy 48GB in the largest model.

File processing is provided courtesy of 1 to 8 X-blades using N+1 clustering, with each blade supporting up to 24GB in the largest configurations.  

File protocol support include the traditional NFS and CIFS, as well as the newer pNFS and its EMC-specific  predecessor, MPFS.

As mentioned before, the back end is a new 6Gbps all-SAS design, supporting combinations of both 3.5" and 2.5" form factors.

Gateway versions of the VNX are still available for customers who want to add file processing capabilities to existing block-oriented environments, most commonly a large VMAX or DMX environment.

Likely Impact Of The VNX and VNXe

Once all the competitive bluster has blown over -- and there will be a great deal of that, to be sure -- the likely outcome is pretty simple: the VNX will come to represent the new standard for mid-tier storage in so many areas.

That includes obvious aspects like performance, simplicity and efficiency -- as well as more subtle areas as data protection, management integration, packaging and the like.

Make no mistake -- the VNX family is a major advance in the intensively competitive mid-tier storage segment.  Competitors will probably need to respond with new designs of their own -- simply refreshing older designs won't be sufficient.

Stepping Back

It should be obvious to most industry observers that EMC continues to invest heavily in the mid-tier storage segment -- not only advancing the successful CLARiiON and Celerra into the new VNX family -- but establishing a beachhead in the new, important and historically under-served segment represented by the VNXe.

Information continues to grow and grow, and storage -- as a segment -- is hotter than ever, attracting strong interest from smaller vendors and more established ones alike.

I, for one, am pleased to see EMC continuing to bring its best game here -- ultimately to the benefit of our customers and partners.  

Game on!

Comments

[JeffS]

Chuck,

Congratulations on the new product line. As a nearly 15 year customer of EMC, I'm constantly amazed at the innovation still happening in the storage space. For me, it's always bittersweet when you announce new product, as it seems I replace my EMC gear just prior to a refresh, forcing me to admire the new technology for several years until I can upgrade again. It happened to me with my first FC4700 (CX line being announced shortly after purchase), it happened again with my CX 500 (CX3 being released within a couple of months of purchase), and now with my shiny new CX4-120...

Oh well, I have no complaints, all of the hardware has been rock solid, I'm loving the new Unisphere interface, and the VMware plug-ins are fantastic. That said, perhaps the stars will align next time (send me a hint) and perhaps I'll get in on the ground floor of a new product.

[Chuck Hollis]

Hi JeffS

I share your bittersweet feelings as well. On one hand, it's great to be able to bring all this great tech to our customers like you. On the other hand, the rapid pace of innovation means that yesterday's dream machine is tomorrow's upgrade target.

We do take a large number of measures to minimize the disruption and extend the value of our customer's investment, but there seems to be no good approach that accomplishes both goals.

Thanks for sharing -- you made my day!

-- Chuck

[Jason]

Does anyone know why the VNX only has a single 6 core CPU instead of dual 6 core?

Jason

[Chuck Hollis]

Jason

I'm not the guy who should be answering that, but -- a bit of clarification -- the VNX uses two controllers, not one. Is your question "per controller"?

-- Chuck

[Jason]

Yes, the question is why does each controller only have a single processor instead of dual processors.

Currently we run into CPU issues on our CX4-960's and we are looking at VNX.

This is my understanding so if I am wrong please let me know.

Jason

[Chuck Hollis]

Hi Jason -- I'll ask one of the VNX literate people to respond here ...

[Rich]

Chuck, is there an answer to Jason's questions?

[Chuck Hollis]

There is, and I asked someone to come back here and share the answer, but I guess no one actually did that.

So let me try ...

Basically, the answer is "they're not the same cores you're comparing". The per-core processing power is much greater in the newer VNX models, thanks to our friends at Intel.

That being said, there are workloads out there that can swamp even a VNX. And for those lucky people, we'd like to talk to them about VMAX.

Sorry for the delay ...

[PowerPath]

Wish lists:

Active/Standby mode for Data Mover in Two Data Movers configuration, so they're standing by for each other.

NetApp cluster mode like limited scale out for NAS protocols, single name space cross cluster.

Optional 10G Ethernet ports, or 4/8Gps FC, or 6Gps SAS ports for hosts connection.