Virtual Compute Environment – Technology Innovations

This is one of a 5 post series on the VCE coalition announcement, what’s in it, and what it means.  You can see all the posts together here.

Ok, knowing my beloved and technophile readership, you will tend to have come to this link, and not the others (that discuss the other parts of the announcement: integrated pre-sales/services/support; Solutions Venture and Investment; Vblock partner ecosystem).   I’d highly encourage you to look at them, as the business elements are as important as the technical pieces.

Then again, we can’t deny who we are, and I am a nerd at heart that does business too :-)

So – without further ado, let’s discuss the technology innovations!   Read on – this one is long so to spare RSS readers, I’ve broken the post up…

OK – start from the principle that ideally, you would have infrastructure as a service (IaaS) whether it’s at your site, or at a service provider.   Why?  Infrastructure as a service is fluid, dynamic, and more efficient than traditional “per application” static models.    We call this an Internal if you’re inside an enterprise datacenter and External Cloud if you’re in a service provider facility.   If you can link them together (which we think is possible), VMware/EMC/Cisco call that a “Private Cloud”.   If the External Cloud is open on the Internet, it’s often called a “Public Cloud”.

So – what are the ingredients for Infrastructure as a Service?

• Scalable infrastructure – ability to start small, and grow big
• Be able to be efficient at every level possible
• Be able to manage the whole thing top to bottom – where “bottom” is the low-level plumbing, and “top” is the business application.  This is important to maintain “fluidity” (which results from end to end provisioning models coupled with over-subscribed “big pool models”) while delivering the SLAs needed (which derives from end-to-end visibility and understanding context and dependencies at any time – and of course providing the detailed chargeback picture).
• Be as “cookie cutter” as possible – you want standardized building blocks, and use virtualization to be able to apply for many compute purposes.

If that’s the list of ingredients, here’s our cake and the recipe – Vblocks.  Vblocks are the answer to “cookie cutter” bullet, and the ingredients (Cisco UCS, EMC mid-range and enterprise storage, VMware vSphere 4) are the answer to the “Scalable infrastructure”, and “efficient at every level possible” bullet.

 

 

Now – the above diagram is actually what a “medium” Vblock looks like.  We call this a Type 1 (targeted for 1000 medium VMs per vBlock, and environments with between 800-3000 VMs).   There is a Type 2 (target is very large enterprise and service providers- with a very large degree of horizontal scaling – for customers with 3000+ VMs).   You can see a 3D model of the Type 2 Vblock here.

Note that the distinction is that a Vblock type 1 scales by adding Vblocks from a storage standpoint, and a Vblock Type 2 scales-out from a storage standpoint – as per my blog post on this “cloud storage architectures” here.  

We’re also finalizing a Type 0 which is the entry level Vblock for customers with less than 800 VMs (and will start very small).   There will be others – but not a TON of others.   The idea is to make this like McDonalds – would you like a combo one, two or three – and do you want fries with that?

That may not sound glorious – but it allows us to define scope down to a point where the amount of variability is still flexible, it’s based off of very efficient, very cost effective components, but can have an end-to-end “software mainframe” model.  

Many customers are off working to create their own Vblocks – often calling it different names, and I’ve heard them all :-)   It’s because fundamentally it’s a good idea, and VMware makes that degree of “horizontal infrastructure standardization” possible.   We want to provide a model where they huge amount of time needed to “home brew” is not needed.  

A customer quote: “we found that reducing our infrastructure variability was the biggest move to making us more flexible for our app owners”.

It also it worth pointing out that this allows (Cisco and EMC) to wring every last ounce out of the configuration – whereas we find that most customers aren’t leveraging all the neat stuff their infrastructure can do.

Some examples – and this is comparing with what we find at most of our customers with more traditional virtualized configurations:

• 30% increase in server utilization (through pushing vSphere 4 further, and denser memory configurations)
• 80% faster dynamic provisioning of storage and server infrastructure (through EMC Ionix UIM, coupled with template-oriented provisioning models with Cisco, VMware, and EMC)
• 40% cost reduction in cabling (fibre / patch cords etc.) and associated labor (through extensive use of 10GbE)
• 50% increase in server density (through everything we do together – so much it’s too long to list)
• 200% increase in VM density (through end-to-end design elements)
• Day to day task automation (through vCenter, UCS Manager and EMC Ionix UIM)
• 30% less power consumption (through everything we do together)
• Minimum of 72 VMs per KW (note that this is a very high VM/power density)

Note that all of the above is before we start to apply the next wave of technology innovations we’ve publicly been discussing (things like solid state, FAST, vStorage APIs for Array Integration, and much much more which are all relatively near term).

That all said – there’s even more to the story. 

A VERY important part is the answer to this bullet from the requirements list:

Be able to manage the whole thing top to bottom – where “bottom” is the low-level plumbing, and “top” is the business application.  This is important to maintain “fluidity” (which results from end to end provisioning models coupled with over-subscribed “big pool models”) while delivering the SLAs needed (which derives from end-to-end visibility and understanding context and dependencies at any time – and of course providing the detailed chargeback picture).

This is where all the EMC Ionix acquisitions and R&D over the last year have been directed.   I love it as a plan comes together :-)

So – first of all – here’s how they integrate in the “big picture” sense:

It’s important that EMC Ionix is trying to be focused.   They are not trying to displace/compete with the big traditional management players, but instead, be so focused on managing the “nearly 100% virtualized, cloud-like provisioned, datacenter, built on Vblock type architectural models”, that in that context, we can compete (ergo not try to replace the existing management toolset).    Our long term view is that in the future, datacenters will have legacy environments, and then these new “islands” that are growing with a core idea of fluidity unlike the legacy.   The legacy will still exist of course.

Each of those blue boxes would be a HUGE post to discuss in detail, but here are the core elements.  I’ll also call out what’s here today, and what’s coming

• VMware vCenter manages the core elements – and nothing does that better. 
• Cisco and EMC continue to have fantastic element managers (things that manage the parts of the VBlock), each of which integrate with vCenter (EMC VM-aware Navisphere as an example)
• EMC Ionix manages the totality, not the elements:
  • EMC Ionix Unified App Stack Management is the role in that stack focused on management of the guest and ESX host (think ESX host profiles on steriods) builds.  It has a strong focus on trying to do this with a focus on efficiency (hence the Fastscale acquisition), but also to have very, very strong compliance and remediation tools.   Today, this is based on EMC Ionix Server Configuration Manager (which you can see here).     We are merging in the core Fastscale elements – which takes what was already a “next generation tool” in Configuresoft – and makes it transformational.
  • EMC Ionix Unified Infrastructure Manager’s (or UIM) role can be summed up as “Vblock management” – in other words – all the parts of the infrastructure that support the vSphere 4 environment.   In Q4, this is through extending UCS Manager’s capabilities to add more, and in the early part of next year, this is through complete end to end management of the full stack (including the storage elements themselves).   EMC Ionix Unified Infrastructure Manager is integral to Vblocks, and the preferred modus of integrated infrastructure management. 
  • EMC Ionix Data Center Insight (or DCI) role is the part of the management stack that connects the bottom (infrastructure) with the top (applications), makes the connections, gives cross-domain visibility and context.

EMC Datacenter Insight shipped right around VMworld 2009 (Sept this year), and in that VMware/Cisco/EMC supersession, I demoed it – including a UCS module that will ship in Dec.   You can see how that all fits together here:

Youtube is notoriously bad for detailed screens – you can download high resolution in WMV or MOV format.   Remember that while EMC Data Center Insight is GA, the UCS elements in the demonstration GA in December.  

But… Let’s focus on the “Vblock” management layer.  To restate the challenge – the goal is to have a thing that makes utility-like management of a Vblock (or more importantly a series of them), including server + LAN/SAN network (UCS manager does this well for one UCS system) + storage itself.   As with all things in the VMware, Cisco, EMC consortium, we know customers need choice – and any one element is replaceable.  The value proposition is that the things we build are so tightly focused, so tightly integrated, that if you are looking at something like this – the integration value is so high it’s nearly irresistable :-)

You might have noticed the call out about the new EMC Ionix element in the stack as part of the VCE announcement: Unified Infrastructure Manager

This is a critical part of what’s important in a Vblock – much more comprehensive end-to-end management.    Unified Infrastructure  UIM isn’t GA yet, but is very close.    EMC Ionix Unified Infrastructure Manager is the preferred Vblocks management model.

The first release of EMC Ionix Unified Infrastructure Manager is focused on extending UCS capabilities and will be released this quarter.   It manages compute + network.   The next version, targeted for early 2010, will extend the capabilities (into the actual storage devices themselves) and use cases (push button site configuration against mass template for things like reconfig for disaster recovery).   That early 2010 UIM release manages compute + network + storage as an integrated package.

There’s so much to this news, I’ll do a followup detailed post on UIM v1.0 shortly, as well as what’s in UIM v2.0.

Exciting times indeed!

If there’s a couple things to take away from this:

• Vblock in one sense is simply putting together the best of breed technologies of VMware, Cisco and EMC.   On it’s own – that would be great!
• In the more important sense – it’s trying to be a new consumption model – where compute, network, and storage all come under an integrated management model.  In fact, they become so integrated, that you don’t think of them as separate entities, but rather as a “virtual compute environment”.  You just know you have enough resources to scale to a certain point, both other wise it’s a pool of pooled compute, network and storage resources – fungible and fluidly useable – bringing a new level of economics both capex (higher density) and opex (pay-as-you-go, and use-on-demand).    The fluidity become manageable because you have the tools to manage pools of pools, and the ability to tie it all together so you can see correlation, context, and dependencies in an integrated fashion.

What do you think?   Are we off our rockers – or is this where you want us to go together?

Comments

[Matt Hopson]

Chad, are there any of the VMware management products in this stack like Chargeback or CapIQ as I am looking at them right now.....or is UIM going to replace them ?
Thanks
Matt