One thing that is becoming clearer as we watch the continuing development of the ecosystems being built on top of virtual machines is the reinvention of system services typically associated with sever operating systems.
VMware with the release of vSphere 4, for example, has started to deploy a range of system services on top of it virtual machine. The basic idea is that these services will be more efficient sitting on top of a virtual machine as opposed trying to invoke existing system services in Window Server 2008 through Microsoft's Hyper-V. This approach is one reason VMware can claim that most applications now run faster on top of VMware vSphere than they do on their native operating systems.
What all this means is that a distributed sets of services evolve, virtual machine software becomes a platform. Today, most IT organization use virtual machines to simply increase the utilization rate of their servers by running more applications per server. As virtual machines evolve into a platform, the services that applications will be evoking will soon be distributed across any number of virtual machines running on any number of servers. In essence, virtual machines will morph into platforms that provide flexible sets of distributed computing services.
No doubt this will create some management challenges. Most of the tools that IT people use today were designed to manage virtual machines more efficiently. But what will be needed are tools that optimize services based on the availability of any number of processors that make up the entire virtual ecosystem.
These distributed set of services, which VMware has positioned as a foundation for cloud computing, are fine in theory. But we've already seen that application owners are little skeptical of virtualization in its current form. Anything more sophisticated is going to require the ability to guarantee specific levels of service for every application owner.
As we all know, service level agreements (SLA) are touchy subjects. One company trying to position itself as a management tool for virtualization services is Fortipshere. Recently appointed Fortisphere CEO Siki Giunta says the existing tools from companies such as VMware, Microsoft and Citrix on the one hand and IBM Tivoli, CA and Symantec on the other are all focused on tactical IT operational users rather than the real service level concerns of the application owners in the business.
The problem is that historically IT organizations have not eagerly volunteered to attach an SLA to any given application. And application owners have not really had the tools available to really measure compliance with any given SLA. A recent study funded by Stratus Technologies found that 46 percent of nearly 300 IT professional surveyed don't review SLAs. Another 11 percent were unsure if they did or not. And 51 percent said they did not know what the value of an hour of system downtime might be for their organization. What all this seems to confirm is that in today's world, far too many SLAs are not worth the paper they are written on.
But as virtualization continues to evolve, measuring the actual compliance with the performance goals set forth in an SLA is going to be the only way applications owners will first gain confidence in virtualization, and then secondarily start to optimize their applications to take advantage of sets of distributed services.
Once that happens, we will have effectively reinvented enterprise computing around highly flexible virtual machine platforms that for the first time will allow IT organizations to respond to any application demand almost instantaneously.
The question is not whether we'll ever get there, but rather how long it will take to achieve it?