Continuing on our post sequence from the past few days, Joe Tucci's last topic -- and perhaps the most major one -- is how he and Paul Maritz firmly believed enterprise IT would evolve over the next few years into a "private cloud" model.
Note: for regular readers of this blog, the ideas presented here are pretty much the same as my previous posts on the same topic.
But it was awfully good to hear Joe and Paul stand in front of a large room of financial and industry analysts, and spend a good hour on the topic -- it's that important.
And when Paul started his presentation following Joe, it smoothly picked up where Joe left off.
Joe put up a slide with four thematic topics (in no particular order):
- data centers that were becoming more and more virtualized
- early forms of cloud computing that were attracting attention and interest
- virtual clients that could deliver a standardized rich user experience on any variety of devices: laptops, cell phones, etc.
- virtual appliances that would deliver applications and infrastructure software in virtual containers with all the needed bits.
But the question remained -- how would these elements come together to define an end-to-end model that would transform enterprise IT?
Data Centers and Clouds
Important attributes like "trusted". The ability for IT and the business to control what happens there.
The fact that most data centers can do what they do pretty reliably.
And, of course, that all-important security attribute.
There are things that aren't ideal about today's data center, but most of us know what those things are, so Joe didn't spend much time on those topics, and neither will I.
On the other hand, there are several important attributes that make various forms of cloud computing potentially attractive.
Applications use exactly what they need -- when they need it -- and no ore -- out of a giant pool of resources.
As a result, this behavior was one of the things that made cloud architectures extremely efficient at what they did -- in addition to doing them at great scale.
Resources in a cloud are always available on-demand -- whether to provision a new application, or to bump up performance if needed. And when the application doesn't need quite so much, the resources are returned to the pool for others to use.
The resulting flexibility of this model was becoming extremely compelling, especially in larger enterprise IT organizations that were running many thousands of workloads, and could benefit from cloud-like attributes.
Combining the Best Of Both Worlds
But how were we going to combine the best of both worlds? How were we going to get to a world where we had an environment that had the trust, control, reliability and security of the data center, but had the dynamic, efficient, on-demand and flexible nature of clouds?
Virtualization was already being widely deployed in data centers around the world, virtualizing them and delivering a great deal of efficiency.
As customers started to exploit VMware features such as large-scale DRS (dynamic resource scheduling), they were -- in fact -- creating gian pools of resources that could be managed as internal clouds, behind the firewall.
Over time, enterprises with multiple data centers would be very interesting in putting more and more of their computing resources into virtualized environments, building larger and larger resource pools that could span data centers before too long.
Virtualization could also be used as a key technology to federate with external providers as well, leveraging their resources for additional capacity, or economic advantage, or geographic location, or whatever.
For this model to work, information would need to follow the workload in such a way that flexing across this giant pool could happen quickly and cost-effectively, even with larger databases.
Security -- controlled by the enterprise -- would be essential to create the trust and confidence required for this model to work.
And, although not shown, management and orchestration tools would be needed by enterprise IT to manage this combined pool of internal and external resources to ensure service delivery regardless of circumstance.
The Private Cloud
The result was a new form of enterprise IT computing -- a "private cloud" -- one that had all the positive attributes of both traditional data centers and computing clouds, and -- better -- it had an obvious migration path from today's application, and would provide the control and security that enterprise IT demanded.
The notion here was the ability (not the requirement!) for internal IT resources to not only federate with each other, but to be able to federate with multiple compatible external service providers as needed.
This "private cloud" model would give IT many of the advantages of a cloud model, and few -- if any -- of the disadvantages of current approaches. As a result, IT would have new choices as to precisely how much their workload they wanted to run internally, or externally, or any dynamic combination.
Virtual applications (both user applications and infrastructure applications) could easily move back and forth between internal and external assets, if wanted -- to either acquire flex resources for a processing surge, or perhaps for business continuity reasons, or perhaps to put specific applications closer to users.
The reason for running a particular application in a particular location really didn't matter; what was delivered was an extremely flexible mechanism for pooling assets that were either owned, rented or both.
As long as the application, along with its information and security container, could be moved easily, yet still controlled by enterprise IT, the model should prove extremely attractive to large enterprise IT users everywhere.
Going a bit farther, a minor evolution in desktop virtualization technology (think vClient) would give enterprise IT users the ability to start thinking in terms of provisioning users, rather than devices.
Regardless of device -- thin client, thick client, desktop, laptop, web browser, smartphone,network connected or not -- rich user application experiences could be provisioned and delivered using the same model -- simple connected to the private cloud.
Stepping Back A Bit
Now, there's more to the picture, of course, to make all of this real.
Management and orchestration technology would have to be very comfortable with this dynamic, network oriented model, not to mention be very familiar with virtual machines and other abstracted IT entities (think EMC Smarts).
And traditional security models and technologies wouldn't work well in a private cloud -- policy would have to follow the application and its information wherever it went. Something that didn't depend on physical location would be needed (think EMC' RSA security division).
But, more importantly -- how would VMware (and virtualization in general) need to evolve to make private clouds real?
And how would the compatible service providers be encouraged to invest in compatible infrastructure, as well as the newer tools required to manage and bill it all?
Those two topics were covered as the presentation evolved.
Next up -- Paul Maritz of VMware.