What do Virtualization and Cloud executives think about 2011? Find out in this VMblog.com series exclusive.
Contributed Article By Patrick Kerpan, chief technology officer (CTO), CohesiveFT
2011 in Cloud Computing - The Empire Strikes Back!
Looking back on my 2010 predictions, I say I went 5 for 10, my colleagues insist it it is 7 for 10, you be the judge. (http://vmblog.com/archive/2009/12/30/2010-in-cloud-computing-game-on.aspx). Where I missed was actually believing that cloud-washing would start winding down (instead of ramping up) and that the basic and simple structures of the cloud market would become evident (unfortunately opacity has increased as some vendors and analysts pursue a "scare the enterprise users" approach to their business ambitions).
(As a note, I am liberal in my cloud definition, so when I say "cloud" I mean public, private, partner, hybrid, etc.. and am talking about Infrastructure as a Service. All standard disclaimers apply. While my views are obviously informed by my work at CohesiveFT, the predictions and commentary below are mine and do not reflect corporate statements from my employer.)
The "hardware support for software encryption" conversation will NOT begin. This is an about face from last year's predictions. No matter to what degree enterprises CANNOT have their data in motion in "plain text" on a 3rd party network over which they have no insight, visibility or control, the hardware side of the technology market will expend resources on security approaches that have the potential for providing customer lock-in, over solutions which would leverage existing customer skills, understanding and IT assets.
"POA" (plain old applications) to the Cloud is the biggest IT migration ever. Enterprise IT lives out its life one business topology at a time. These topologies range from several servers to several 10's of servers, are fairly tightly coupled, and for the most part don't interact with other business topologies. These are "plain old applications". Helping enterprises embark on the long, slow migration of these systems to agile infrastructure is the big payoff. "Cloud" is becoming the default design center for the enterprise, and as such is the biggest migration exercise ever, bigger than the Y2K.
The cloud approach begins to show profound implications in how enterprises organize their IT functions. The cloud data center is becoming a three-layer architecture with three distinct planes of operational control. The bottom layer is the physical layer. Its job is to be fast, fat, and flat. It runs state of the art hardware that can be multiplexed by the middle layer (the virtual infra layer) to make the physical virtual. The virtual infra layer has a pretty focused job too; get virtualized compute and storage into play for users of the infra, and be easy to own, manage and operate for the provider of the infra. The top layer, is the "user-cloud" layer. It is from this perspective that individuals, IT operations and lines of business deploy business capabilities running on top of the virtual infrastructure layer. These three layers have distinct interfaces, and a clean separation of organizational concerns. This allows enterprise IT a dramatically new framework for managing expertise, staff, service partners and software and hardware vendors.
Hypervisor vendors will continue to struggle with the return of the server operating system. After years of providing a "thin layer of software called a hypervisor" to enable the running of virtual machines, now it's back to operating systems. All current Linux server operating systems can run virtual machines, and of course Windows Server 2008 can run virtual machines. As these incumbent vendors began targeting the server virtualization market, think back to the phrase that was used. You were told you didn't need a server operating system any more, rather, and in their words, "you install a thin layer of software called a hypervisor". I always loved the turn of phrase "called a hypervisor", not "is a hypervisor". This was a critical piece of positioning that helped this market take off. Alternatively, if customer's had known they were installing a custom Linux OS distro from a virtualization vendor, things would have been quite different.
Savvy enterprise buyers will tell their hypervisor vendors to "shut up and drive". Like a nosy taxi cab driver or limo driver, hypervisor vendors are involving themselves in additional feature and function areas in ways which disrupt and deform the powerful new separation of concerns in the enterprise cloud datacenter. As hypervisor price commoditization has occurred, the providers of hypervisors are looking for ways of returning that "not quite so thin anymore layer of software" to a position of power. In the interim they have focused on making pools of hypervisors easier to own, manage and operate. This is a good thing! They are awakening to the critical need for federation (making pools of datacenters of hypervisors easier to own, manage and operate). This is good too! However, they are also trying to make the tightly controlled hypervisor a nexus of control over their customers. With traditional server software - customers have complete control of what they do and don't run in the server OS. Enterprises can make it as lean or as fat as works for their organization. Hypervisor vendors control their stripped down OS that runs on customer hardware - and they make decisions about what can run in it. And now hypervisor vendors are telling customers they (the hypervisor vendor) will make all sorts of single tenant, single choice, unilateral decisions about what you need in that layer of your infrastructure. Savvy buyers - beware.
"Cost Zombies" start to be a concern. Many a virtual machine "sprawl management" startup has come and gone over the last few years, and as a rule I still think we don't have enough virtual machines out there. But it is possible now to go "ouch, $500 this month on some cloud virtual machines I had forgotten about". As enterprises become more facile with agile infrastructure expect them to demand time and dollar budgets to be part of the cloud server launch parameters. This could be provided and enforced by either your virtual infra provider or your user-cloud providers. As invoicing and purchase orders join the credit card in the public cloud offerings, it will be interesting to see the emergence of these strategies.
"Federation" becomes a big part of the enterprise virtual infrastructure conversation. Providers of virtual infrastructure such as IBM, VMware, Citrix, and Eucalyptus obviously have intellectually understood how many datacenters a large enterprise has. This year begins the emotional and visceral understanding of what this means and how they need to appear as continuous fabrics as opposed to islands of hypervisor and virtual machine isolation.
Microsoft has an "aha" moment around Amazon-style services via the Azure "VM role". Azure has fought a retreat since its inception, resisting being an infrastructure as a service offering that can run X86 workloads defined, controlled, manufactured and uploaded by customers. It began as a "language runtime cloud" offering a set of platform services which could be consumed by your .NET byte code and executed "in the cloud". Virtual machines had no role. Unfortunately, and somewhat obviously, enterprises couldn't even begin to imagine the attack vectors of such an approach much less envision the defense, governance and audit approaches necessary for it. Then came the "well, it is a virtual machine, and you can pay-per-hour for a little, middle or big one, upload your .NET code, control some parts of it, but no you cannot fully control it as a virtual machine" offering. There still were not a lot of takers for that approach. Now they have announced the Azure "VM Role" which allows customers to upload their pre-manufactured Windows Server Virtual Machines and run them. Customer's have wanted and waited for this - and with the investment Microsoft has made in its Azure datacenters - customers will use this. Customer acceptance of this will light up the bulb over Microsoft's head and we should see some good come of it. And we hope they will bite the bullet and announce/deliver support for Linux devices in the VM Role as well.
HP will NOT emerge as a visible super scale cloud provider. For Infrastructure as a Service, Amazon remains the only super scale player to date. HP and its billions in capital will announce more and more clouds and cloud services - but won't provide a super scale infrastructure that gives you easy, API-driven use of virtualized containers. They will announce wondrous, magical things in press releases, complete with quotes from customers - but somehow the rest of us won't be able to see it or use it.
Another public cloud offering will reach critical mass and exit 2011 poised to be "the enterprise public cloud" in 2012, and Amazon's first serious competitor. There are only a handful of possible participants from the enterprise computing space, a couple from the Internet consumer space, and some dark horse telcos around the world. I think one of them will put it all together sufficiently to go toe-to-toe with Amazon, but with an enterprise focus. At least one of these players (my money is on IBM), will get through the business model complexities of cannibalization and sales compensation in 2011 and be strongly positioned as the "enterprise grade" public cloud. If not, then Amazon will have a free pass to become the dominating force of both the Web 2.0 and the Enterprise cloud markets.
We will begin to speculate about the first Cloud IPO. The name most mentioned in this topic will be RightScale. Fair disclosure, my company, CohesiveFT, partners with RightScale in a number of accounts, competes in other accounts, and are a non-intersecting set in others given differences in feature set and target customers. That said, from a financing and valuation metric (if one believes the rumor mill), they appear to be the leader of all the private companies/startups in the entire cloud stack. As the recession recedes, and the cloud market matures, they have a real shot to be the first big investment success of the cloud era.
About the Author
Patrick Kerpan is the chief technology officer (CTO) for CohesiveFT, the provider of the industry's first Secure Application Container for Cloud Computing; offering integrated image automation, virtual networking, and topology automation. This "user-cloud" solution provides enterprises the capabilities needed to achieve their cloud objectives, meeting the challenge of the largest IT migration since the Y2K. In this role, Kerpan is responsible for directing product and technology strategy.
Keep up with Pat at http://blog.elasticserver.com/, http://www.twitter.com/elasticserver, and http://twitter.com/pjktech