Today I’m interviewing Matt Coviello, Director of EMC Demo Strategy, who has built and run an internal private cloud for a specific use case. His perspective and experiences can be a useful insight into one of the many ways private cloud concepts can be used to solve specific business problems.
Hi Matt – thanks for sharing some time with us. Can you tell us a bit on how all this came to be?
Sure. A few years back, we were handed a very specific business problem. EMC’s presales organization (~3000 people) continually conducts demonstrations in front of customers. Doing this the traditional way meant either assembling the demo in front of the customer, or perhaps traveling to a location built for that purpose. Neither was ideal, if you think about it.
We wanted to create an on-demand service that allowed any EMC presales person to quickly create a custom demo on demand. It had to be zero-touch, very dynamic and – of course – cost efficient. Back when we started this, we called it a “demo cloud”, but given modern terminology, it’s got all the aspects of a private cloud.
So, what was the justification behind all of this?
As we started this effort, the primary motivation was reduction of expenses compared to how we were doing things in a more traditional manner, but – as we got into the project – it became clear that the business benefit – enabling our presales people to provide a better customer experience much faster – was the dominant motivation.
Put differently, we shifted our focus from cost reduction to business enablement – and ended up delivering enormous cost reduction along the way as well. We are now expanding the scope of the “demo cloud” to be the key enabling technology behind the Global Demo Program for EMC run by my group out of Global Solutions.
How does the environment work?
It’s all self-service. EMC people use a portal to request a pre-fabricated instance of a software environment or a specific use case, such as demonstrating email archiving. The environment can then be customized by the presales person to incorporate unique aspects of interest to the customer – scripts, interfaces, etc.
It’s all real-word stuff, and not some synthetic demo.
As EMC continually offers new products and services, we’ve also got a nice methodology to quickly gen up new environments, publish them quickly, and get feedback on the ones that are out there.
So, how big is this environment?
Today, we’re running ~900 virtual machines, 75 TB of storage and approximately 2.5 TB of server RAM. We need to roughly triple the size of the environment, and we’ll be using the new Vblock approach for that.
I found it interesting that you called out the RAM number – why?
I think most people who’ve done VMware at scale realize that you usually run out of server memory before anything else. One of the things that intrigues us about Cisco’s UCS is the ability to run 4x memory at far lower cost than we’ve been able to achieve before.
More memory = more virtual machines for less cost. It’s something we really care about.
What about manageability?
For a long time, we ran the operational side of the environment with 1.5 headcount. That was a little too lean, so right now we’ve grown to 3 people, anticipating a dramatic growth of this environment.
Growth? How?
Well, we’ve got the normal organic growth associated with the demo use case. It’s the normal adoption curve, and it’s taken off for us. If you make customized demos easier to do, you’ll do more of them – and hopefully sell more stuff. It’s a high-class problem to have!
But, beyond that, our presales people need on-demand computing for a variety of other purposes as well – including doing extensive performance analysis, customized scripting and integration work – you name it, they’ve got a bottomless appetite for infrastructure-as-a-service.
And I think everyone understands the link between what we’re doing here, and overall revenue generation for the company.
What about the bugaboos of management, backup, security, etc.?
Well, one thing we’ve got going for us is that we don’t need the very highest levels of availability and security for this use case. I mean, the technology is there for us if we need it, but – so far – we haven’t really needed to make this environment bulletproof in a traditional IT sense. That being said, it’s been very robust and stable for us.
The one thing we’ll never compromise on is our operational model. We’re fielding an enormous amount of IT business value with a very small investment, especially when it comes to people. Adding people can make you slower and less nimble, not to mention more expensive.
You did this outside of a traditional IT framework – why?
Back when we started this, we really didn’t have the skills and methodologies within our IT organization to get us where we wanted to be fast enough. Were I doing it over again, we’d certainly be leveraging their skills and capabilities.
I think this speaks clearly to the observation that there’s a new way of doing IT: this stuff is built differently, operated differently and consumed differently.
Lots of interest out there in things like chargeback and billing – how does this work for you?
At EMC, this entire environment is used exclusively by our presales organization. Although we’re responsible for the normal balanced scorecard, it’s all paid for by one part of the organization. So this wasn’t a problem for us.
Now, if we decide to use this environment for other parts of the business, we’d probably have to come up with the next level of cost allocation.
What about external service providers?
You know, that’s always an option. We’ve built the environment in such a way that we can do that quickly if we need to.
We continually compare our internal cost numbers to external options, and we seem to be ahead of what the service providers can do today. That could change, though.
And, if sometime down the road, we need an increased presence in, say, China or India, we’d probably consider the service provider option, although controlled and managed by us.
What’s ahead for you?
Well, as I mentioned before, we’ve got to take our capabilities up to the 4,000-6,000 VM range over the next few quarters. We’ll be using the Vblock approach to do that – we’re approaching that scale, and can justify that much infrastructure. And I’ve got to go find some budget to do that 
We’d like to implement a number of new consumption portals for different use cases within our presales organization – lots of good stuff coming from VMware and EMC in that regard. And we’ll also be implementing the new EMC Ionix stack as well.
Just because you work for EMC, does that constrain your choices?
I know what you’re thinking, but it’s really not that way. Earlier iterations of this environment were built with a mixture of EMC and non-EMC technology. We had a job to get done, and we picked the best tools for the job.
But now, we’ve got some exceptional tools coming out of VMware, Cisco and EMC – they look like they were designed with my environment in mind. And we’d be foolish not to take a hard look at them. Being a part of the EMC Global Solutions organization now puts us in a perfect spot to leverage the Vblock and other new technologies coming out of EMC and its partners to scale our cloud in the most cost effective manner.
Any other final thoughts?
Needless to say, we’re pretty proud of what we’ve done here. It might not sound like much, but it’s been a lot of hard work, and it’s delivered enormous business value that isn’t questioned by anyone.
I’m looking forward to being able to share our experiences with others who might be interested in doing the same thing.
Based on what I see, there are going to be many projects like this going forward.
Trackbacks /
Pingbacks