In previous blog posts, I've discussed in some length how EMC's own IT organization is progressing in their journey to the private cloud, running through the phases we're using as a reference model.
Well, progress appears to be a bit ahead of schedule. Serious planning work is now being done to offer a broad range of IT-as-a-service to the EMC organization. And I think there are some useful insights for everyone here.
Full disclaimer: these slides are freshly pilfered from Sanjay Mirchandani, EMC's CIO. Sanjay likes to give me a tough time that I steal all his material. Sanjay, I'm not actually stealing your slides -- I'm only just liberating them. Rest assured, I only steal from the best :-)
What's This All About?
Although EMC's IT organization supports a world-class technology company, they're faced with many of the same challenges as our customers: aging platforms, compelling motivation to be both more efficient and agile at the same time, and so on.
As a result, they've gone "all in" towards transforming EMC's considerable IT operations into a private cloud model: fully virtualized infrastructure and applications delivering IT as a service using a federation of internal and external resources.
Sure, we'd like to showcase EMC's products and technologies. But there's more to it than that: EMC -- like many companies -- needs a competitive advantage from what IT can do. IT proficiency is an important part of EMC's overall business strategy.
Documenting their challenges and successes along the way has been incredibly valuable to many audiences, both internal and external: from technology requirements to integration to operations to organizational changes to entirely new relationships with the business -- there's been plenty to talk about.
The Three Phase Model
We're using a three-phase model to describe the journey.
The first phase you're probably familiar with -- it's virtualizing the applications that IT owns -- the less business-critical stuff. New technologies are introduced: hypervisors, pooled storage, etc. -- but there's not much process change.
Most customers operating at scale understand the enormous economic benefits that result from consolidating server and storage infrastructure.
The second phase is a bit harder -- it's virtualizing the business-critical applications. These are the ones that people
really care about: email and collaboration, big databases, customer-facing apps -- the big hairy ones. Here we're talking about replacing the processes -- and supporting technologies -- that go right to the core of what IT does for the business.
New ways of provisioning and managing service levels. New ways of protecting data. New ways of securing and ensuring compliance. Not only are the technologies changing, but the processes and the roles of the people managing those processes as well.
As a result, while there are plenty of technology issues to go sort out (isn't that always the case?) the real challenge has been in the maturation of the supporting processes -- IT is now run differently than before.
The third phase -- delivering IT as a service -- may seem like some distant and unobtainable nirvana to many. The hardest part? You're changing the fundamental relationship between IT and the business towards an internal service provider.
The surprise to many is that this goal is now within reach: in addition to EMC IT -- we're working with many customers who see this as an achievable outcome during 2011.
And -- frankly -- it's really exciting to see how this is all unfolding.
Tackling Phase 2
Phase 1 -- the straightforward virtualization of non-critical applications that IT owns -- is ground that many IT organizations have either done, or are actively doing today.
We get some questions here and there, but not a lot -- generally speaking.
By comparison, the interest about how EMC IT has moved through Phase 2 -- virtualizing critical applications -- has been considerable.
Our IT group has been continually besieged with requests to share how they've moved two particular key functions to the private cloud: large-scale email, and our enormous Oracle eBusiness landscape.
For example, EMC's Exchange 2007 environment is not trivial: we're talking ~50k mailboxes and ~20k mobile users that handle an average 4.5 million messages per day. And -- as users -- we all not only expect it to be available 24x7, we expect sub-second response as well -- even when we're emailing a 20MB powerpoint to dozens of people.
It now all runs nicely under VMware, and includes all the archiving / security / DLP bells and whistles as well -- mostly built from EMC technologies.
Even more interest is generated from our progress in virtualizing our ginormous Oracle eBusiness suite. It's the engine that runs EMC's transactional business using a complex landscale of 67 application servers.
Needless to say, performance and availability is *critical*: 40k named users that can process 25 million transactions per day.
And yes, it's mostly virtualized -- except for a few particularly hairy Oracle RAC databases. Soon, soon.
It's one of those IT stories we all like to hear about: huge cost savings, better service levels, better security and data protection, faster and more responsive to business needs, far easier to upgrade and enhance. I'm hoping we do an extended white paper on this before too long.
Most importantly, these stories create clear proof and additional confidence that -- yes -- you can virtualize business-critical applications. And create a huge win for the business in the process.
And On To Phase 3
I've given you a preview before on how EMC IT wants to offer a full suite of IT-as-a-service offerings back to the business, as well as other portions of IT. At a high level, we're talking four categories of "aaS" -- infrastructure, platform, software and applications, "desktop" -- all backed by integrated infrastructure and process.
The vision has become a plan, and work has begun in earnest. The stack has been defined, and it makes wonderful sense -- at least in the context of EMC's business model.
At the foundation is infrastructure: compute, network, storage and backup -- all nicely packaged in a Vblock (naturally!).
Storage, in particular, is fully virtualized, automatically tiered and progressively pooled over larger geographies using technologies such as FAST and VPLEX.
And, of course, all the backup and replication is deduped -- using Avamar, Data Domain and RecoverPoint -- with DPA (data protection advisor) managing the service.
No surprises here.
It starts getting more interesting when we get to the PaaS (platform as a service) layer. Underpinning the platform is a data management layer. We're explicitly calling out Oracle (lots of that), SQLserver and similar convenient-to-use databases, Greenplum for on-demand data warehousing and large-scale analytics, and an open slot for something we'll be discussing in the future. Naturally, all of this runs on fully virtualized physical infrastructure.
One hint: over time, we'll likely be talking more about "data virtualization" -- the ability to conveniently use data and meta-data regardless of the underlying data management approach.
On top of that, you'll see some of the usual application-oriented components: development tools, frameworks, integration and runtime management using Springsource, security and GRC components sourced from RSA, enterprise content management and ILM (information lifecycle management) using Documentum, SourceOne and other EMC IIG products, and a few other bits.
Once again, the idea is to expose these as services -- either individually, or in combination, to anyone at EMC either building applications, or leveraging off-the-shelf packaged software.
Going up a bit farther, EMC IT plans to offer distinct SaaS offerings for critical applications used across multiple business units. For example, a significant portion of EMC's information inventory has to do with products, customers and transactions, hence the need for MDM (master data management) as a service -- independently of how the data might be used.
ERP and CRM are logical additions, of course. High-level GRC management has been explicitly called out as an "as a service" capability, as has on-demand business analytics. These on-demand application services will complement the more traditional application delivery models -- both categories will use the underlying fully virtualized "aaS" infrastructure and processes.
Capping the stack is a block labeled "desktop as a service". Perhaps a more accurate nomenclature would be "managed user experience as a service" although that can be a mouthful. That project has just finished a successful initial rollout, and as a result we'll be migrating user groups at full speed during the remainder of 2010 and throughout 2011.
The big insights here? Easy -- a clean and digestible picture that describes EMC's functional IT architecture in a single slide. Sure, there's plenty of supporting detail missing, but this at-a-glance view helps anyone quickly understand the end-state from a non-IT perspective.
Does your team have a slide that looks like this?
From Projects To Capabilities
Earlier, we shared the key project initiatives that EMC IT was tackling in support of this transition. Some of the projects got done during phase 1 (virtualizing IT-owned applications). Other, more difficult projects got done during phase 2 (virtualizing business-critical applications).
Elements of both are now forming the key underpinning of phase 3 -- IT as a service.
On the left side of the chart are the key projects -- for example, establishing a CMDB and using it consistently.
On the right side of the chart is the resulting capability that will be needed -- for example, integrated management and automation.
A key insight for me? Note that the language has changed from specific technologies and projects to infrastructure-wide capabilities that are needed across a broad range of use cases.
And as these capabilities broaden and mature, more of the workload will be brought over to the "new IT", and progressively less of it will be run on the "old IT".
Getting There Is Half The Fun
Now, it goes without saying that all of this won't happen at the same time. Certain capabilities will manifest themselves earlier in the journey; others will take longer to mature. How best to manage this process?
The EMC IT team has proposed a "quarterly turn of the crank" to incrementally make new "*aaS" capabilities available to the business.
Every quarter, each layer in the stack will grow new capabilities in addition to existing ones. No big bang, no gut-wrenching transition, no waiting for things to be perfect, etc.
Here's a graphical example of how the infrastructure-as-a-service will likely play out from quarter to quarter over the next
year or so. You can see the incremental progression from fairly simply infrastructure services to far more robust and complex ones.
And, like all roadmaps, they're subject to change depending on how things progress. What doesn't show here are the growing feedback loops from the users of IaaS that have started to weigh in on their priorities and specific use cases -- just like the interaction between a service provider and their customers.
The New Business Of IT?
Squint hard, and you can see the beginnings of a service provider business model starting to take shape.
Proficient IT users who need specific IT services.
An infrastructure and set of capabilities that can progressively deliver these services more efficiently and more specifically over time.
A marketplace of virtual pricing that will help prioritize the effort and investment.
And so on.
It's one thing to make the broad industry claim that -- over time -- most IT organizations will begin to look like internal service providers.
And it's another thing entirely to watch it happen before your eyes.