Virtual Architect

SCOPE Your Infrastructure

Targeting the right servers to virtualize can be challenging the first time around. Using SCOPE can make that process easier and more efficient.

• By Danielle Ruest and Nelson Ruest
• 01/01/2009

Ideally, the answer will be every single one of your production servers. The dynamic data center is one where all user-facing resources are virtualized and can be moved from one host server to another to ensure the very highest levels of service. However, you have to start somewhere, and identifying which servers to start with can be challenging.

In production, you'll want to get the best performance from both hosts and virtual machines (VMs). After all, even if your main goal is to improve physical resource utilization, you'll still want to make sure all systems perform at their best. That's why you'll need to look at each virtual service offering, or each workload you'll be virtualizing, one by one.

Where to Begin
The best way to do this is with a structured approach. Begin by performing a comprehensive server inventory in which you identify resource usage, server workload, application configurations and more for each server. Some of the tools for performing this analysis were discussed in the September/October issue ("Roundup: Capacity Planning Tools") and in the July/August Virtual Architect column ("Which Servers Should You Virtualize?"). Once you have the data, look at the workloads themselves to determine where to start.

One good way to do this is to rely on SCOPE, a conceptual tool shown in Figure 1. SCOPE lets you quickly identify which workloads to start with. Begin in the lab-maybe even with the lab machines themselves-and slowly move your virtual workloads into production.

[Click on image for larger view.]

Figure 1. The SCOPE Curve helps determine which servers to virtualize first.

SCOPE is simple to use. It begins with the easiest workloads to convert, then moves upward through the most complex and critical to your business. It divides workloads as follows:

-
• systems such as Dynamic Host Configuration Protocol (DHCP) servers. Web servers are ideal in this category because they often run stateless workloads or workloads that are Read-only.
• Productivity workloads include systems that run printers, file shares, collaboration services such as Office SharePoint Server and other servers that act as front-ends to n-tier applications. Also included is any system used to share applications centrally, such as Terminal Servers.
• Operational workloads include any system designed to manage your infrastructure. These are often complex implementations, but are still relatively easy to migrate because they often include the capability of transferring a role from one server to another, making the transition from physical to virtual-known as P2V-simpler.
• Complex workloads include systems that run high-availability or load-balancing services, as well as those that have high I/O throughputs, such as databases. If a system is load balanced, migration is simplified because you can add new VMs with the role and integrate them into the load-balancing cluster. When enough machines have been virtualized, you can decommission the physical nodes. Clustered machines also use this model. For example, with Windows Server 2008, you must migrate clusters from previous versions of Windows Server. Migrating to a cluster made of VMs is just as easy as migrating to a physical cluster. High I/O workloads present a trickier migration, but by the time you get to these workloads, you should be very familiar with the capability of your virtualization infrastructure and with VMs in general.
• Special workloads are reserved for last, giving you the opportunity to gain experience in the process. These workloads are usually fewer in number and will include items such as business-critical, manufacturing and legacy apps. Here you have to look to special procedures for virtualizing the workload to make sure it still runs properly once converted.

End Goal
At some point in time, you'll hit the value to business barometer-the point at which conversion takes more effort than normal and you begin to wonder if it's worthwhile. We would argue that conversion of 100 percent of your infrastructure is the goal. Once all systems are converted, you'll have a truly dynamic data center; one where all end user-facing workloads are virtualized, and the only role hardware has in your network is to act as host servers running virtual workloads.