Wednesday, May 20, 2009

IT efficiency has two components; reducing capital and operational expenditures CAPEX and OPEX). In server virtualization this can be translated to reducing the number of physical servers that need to be purchased as a result of the project and increasing the efficiency of the staff working on those servers.

Most server virtualization projects can measure a distinct CAPEX savings for the project. This can be represented as simply as how many servers did not have to be ordered this year, resulting in immediate dollar savings, as well as how many servers could be decommissioned resulting in long term power savings.

OPEX on the other hand is a different story. Server virtualization in and of itself allows an IT department to put more servers onto a single physical server. These virtual servers by definition are distinct standalone servers that share the same physical space. While one of the reasons that server virtualization works so well is the stability brought about by this separation, the down side is that there is a limited increase in operational efficiency to keep pace with the increase in capital efficiency.

Operational efficiency may actually become worse with server virtualization. Before server virtualization, every request for a server required careful consideration, planning and justification. There was a process. With server virtualization however, other than the administrator time actually required to provision the server, servers can be added without much consideration and as a result the number of servers in the environment can grow out of control rapidly.

According to a recent IDC report the compounded annual growth rate of virtual machines (VMs) is over 40% and predicts that by 2010 there will be over 7.9 million virtual machines. This presents two challenges to the virtual IT infrastructure. First consumers of those virtual machines have to wait for their deployment and second an administrator needs to deploy all those virtual machines.

Today, most environments are a single hypervisor, with the advent of Microsoft Hyper-V and growing interest in Citrix XenServer most environments will have multiple hypervisor or at least will want the flexibility to implement multiple hypervisors with out significantly increasing administrator overhead. If the administrators have to learn and provision each environment separately this will limit the adoption of these newer hypervisors.

What server virtualization needs is a set of tools that compliment the server virtualization project by enabling the administrators to become more efficient. Companies like Vizioncore are creating tools that will improve the provisioning process. Their application vControl for example, allows administrators to manage more VM’s, and automating as many as possible of the manual tasks relating to the care and feeding of the virtual infrastructure.

The first step is to provide a method for the virtualization administrator to regain control over the server provisioning process. Part of the problem in server provisioning are the administrators themselves. They become the apparent bottleneck in VM deployment, but in actuality what slows the process down is that the administrator needs specific information from the virtual server consumer as to how they need the VM configured and deployed. In most data centers this leads to a series of back and forth emails with questions that require answers.

This email chatter takes time, in many cases days and invariably there is always an unexpected question or two that needed to be asked but weren’t. This leads to problems during deployment or worse, in production. Experienced administrators learn quickly to compensate by over provisioning the virtual machine configuration which leads to wasted resources which in turn eat into CAPEX savings.

The solution is a self service provisioning system for virtual machines similar to the one we described for storage in our article, "Virtual Storage's Next Step", that would automate the configuration and asking of the necessary use case questions through a series of customizable VM templates. These templates could then lead to a workflow that would build and deploy the VM automatically under the watchful eye and approval process of the administrator.

Ideally this tool would be browser based so the requester of the virtual machine would not have to learn a specific virtual environment’s interface and the tool should be able to communicate with the major virtual infrastructure providers like VMware, Hyper-V, XEN and Solaris Zones. The user could then simply answer a series of questions about their virtual machine needs and what hypervisor they would prefer that it run.

Once complete the request would be sent to the virtual infrastructure (VI) administrator for review and then if available a workflow manager would then, based on admin approval, create and deploy the virtual machine. Just having the template in place can increase VI admin productivity by reducing the email chatter. However, the ability to automate the building and deployment of the virtual machine is critical to increasing VI admin productivity. Doing so is obviously faster than doing so manually and it reduces the chances for human error during creation.

In our next article we will discuss how virtual machine management and workflow tools can further improve the OpEx savings of the virtual infrastructure and how its integration into provisioning is critical for optimal IT efficiency.