Virtualization design-stage mistakes

Virtualization is still in its infancy. It is estimated that 14 percent of servers in data centers are running on some sort of virtualization and only 2 percent of the work done in enterprise computing is in the cloud. However, virtualization is becoming a leading issue for its main benefit of huge cost reduction through resource sharing.

Desktop images
While virtualization designs are far from maturity, some common mistakes are coming to light. The most common desktop design-stage mistake is that enterprise IT managers end up providing so many virtual desktop images that a major virtualization advantage of simplifying management gets nullified. The number of virtual desktop images typically runs into hundreds and at times even thousands. Are so many images really required? This is a critical question to be kept in mind at the design start. A small group of users are at times given say 10 images just because some people in the group use unique applications not used by others. Designers need to examine what would be more cost effective: giving these people desktop images or delivering unique applications through application virtualization. In most cases the second alternative should work better even if initially the first alternative is cheaper. Over a period, the cost of desktop management of higher number of images is found to exceed the cost of less number of images with more of application virtualization. Just two or three images could suffice with application virtualization instead of say 10 without it.

Planning for storms in desktop virtualization architecture at the design stage itself is another important issue. The designer needs to study user patterns for designing a desktop virtualization. The boot up storm should not overwhelm the environment. The number of hosted virtual desktops directly impacts the hypervisor, the storage, and the network infrastructure. A scenario wherein people work from morning till evening will have different design considerations from the one wherein people work 24/7 across multiple shifts, directly impacting XenDesktop farm design and scalability with respect to boot up storms and logon storms. For instance, in the first scenario above a boot up storm would not impact other users as no users were online before the working day began; but in the second scenario, there were active users all the time. Keeping in mind different user pattern scenarios during a desktop virtualization design is at times overlooked.

Management of cloud computing resources for intelligent workload management is another challenge that has only just been realized. Companies are coming out with products that help in this task. Right Scale offers products to create and manage resources in the cloud, enabling the designer write applications that can automatically scale up and down based on the workload. Virtualization vendors, especially VMware, Oracle, Novell, Microsoft, and Citrix, are working on ways for automating virtualization on private data centers. Red Hat has taken the lead in a harmonization endeavour called Delta Cloud that will provide a common, unified API for controlling cloud resources.

Capturing dependencies
Besides managing to scale resources to match the workload, designers need to reckon two more elements for the best results. These are the ability to set up and configure an application for the cloud and ensure that the data reaches its destination. Setting up an application environment to execute a workload competently in the first attempt itself has been found to be a great design challenge. Currently quite a few number of attempts are required to fix problems before getting the required result. rPath is focussing on this problem. It claims to have captured almost all dependencies involved in setting up an application. FluidOps and Right Scale are other companies redressing these two problems. In the next few months the designer should have sufficient tools to tackle virtualization design-stage problems at the start.