Cloud Computing was a big topic at this year´s EMC World in Orlando. I think it is a given that virtualization is a pre requirement for any kind of Cloud concept that can be implemented today and thus EMC is playing a vital role in the Cloud space with VMware. Following the “keynote” on Cloud Computing and virtualization on Tuesday showed quite well what EMC expects. First of all I want to mention, that I thought it a pretty neat idea to turn a keynote into a panel discussion, because that demonstrates the impact Cloud Computing has on IT – it touches every part of IT and thus every part of a major vendor like EMC gets involved.
The discussion clearly showed that EMC is thinking of what can be done with the Cloud today rather than proposing the overall concept and waiting for it to be technically possible. For EMC Clouds have to tackle the space of legacy applications rather than requiring the users to rewrite all their software. In my opinion this is the absolutely logical step and therefore I liked the content showing how different concepts at EMC support making today´s applications “Cloud ready”. The biggest step into this direction is VMware´s latest release of vSphere that enables outtasking of compute resources on demand while turning the hardware available internally into a resource pool that can be allocated flexibly and automatically. This is supported by resource and system management software as well as storage. It is all done by adding management capacities and predefined behavior to the virtualization capabilities already available and by bringing other components closer to the virtual world by adding direct control over hardware though interfaces to the management program driving and allocating the virtual resources. This concept creates a resource pool out of all involved components (storage, network, compute) that can be allocated dynamically. The feature that tops up the concept is the ability to externalize such services if peak loads require additional resources.
By simulating the environments we are using today and bringing this simulation into such an dynamic space a pre Cloud becomes reality very quickly. This is what can realistically be done today and this is what makes Cloud concepts available to legacy application, short term project requirements as well as test and integration environments.
For my taste the fact that the Cloud concept would require the reprogramming of all software was a little overdone. Yes, I too believe that there is no way dynamic parallelization of computation cannot be reached unless you write programs for such a kind of super dynamic scheduler (like Google does).But this is where computing is headed in the long run. To try and reimplement everything on the spot is absolutely unrealistic and therefore the concepts of bringing at least some of the benefits of Cloud concepts to today´s applications and architects is great. But to say that reimplementation can be avoided in a very long term perspective is just incorrect. I think we should have learned something from the immense cost generated by maintaining the big monolithic legacy apps we do rely on today (If you want something, you make something new because changing the old think to look new will create more cost through maintenance in the long run).
Last but not least the EMC team emphasized many times, that the VMware approach created much less dependencies for customers than giving their applications into the proprietary domains of Google App Engine (where your program only runs with the Google API) or Amazon EC2 (where the virtual machine itself is hard to retrieve once deployed). This is a valid point. And despite the hype created around Amazon EC2 or Google App Engine this addresses the fear of many business users. On the other hand one should state too that EMC as well is building features into their “Cloud OS” that make a customer “want” to use EMC hardware and other EMC preferred services. All in all EMC is doing a good job of opening up the specs and standards for these kind of dependencies enabling other providers to step into the world of VMware and be just as well integrated.
In the long run I am sure standards for Cloud machine images, templates and Cloud programming interfaces will evolve. I think this will be an evolutionary process rather than the job of a standardization committee, because the Cloud idea spreads so quickly and many many different concepts are being tried out every day. Survival of the fittest is not the worst thing to happen here.
As one should save the best for last I can say now that EMC management and engineers obviously understand the need for more effective automation technologies. The discussion returned to the point that such very dynamic infrastructures and environments simply can no longer be managed manually and that the current toolsets available in system and resource management will have to take major steps towards actually automating the maintenance process fully. So in the eyes of EMC and VMWare management and engineering the operational auto pilot discussed in this blog many times and actually implemented in the aAE (arago Automation Engine) is not just a good way for cost cutting or freeing up resources for innovation and change, but becomes an absolute necessity in a dynamic environment where the speed of change is too high to be reflected in human experience. Thus I conclude that the idea of preserving these experiences within an automation engine as described before is the best way to protect investment into these experiences.