How to Virtualize

Data center virtualization projects
Executive Summary
Data center virtualization projects have a ripple effect throughout an organization’s infrastructure and operations. Collapsing multiple physical devices into software affects the people and processes supporting the data center. Before embarking on data center virtualization deployments, enterprises should account for how business processes, administrative rights, capacity planning, performance monitoring tools and security strategies will need to change.

Introduction

Virtualization projects are under consideration or already in progress at many organizations looking to decrease their capital and real estate expenses and to launch energy-conservation initiatives. Through the use of special virtualization products and services, data center managers can apportion computing hardware resources to operating systems and applications on a time-sharing basis. In doing so, they can significantly boost the utilization and efficiency of servers, network equipment and storage devices. Such projects reduce the overall number of physical devices needed, the floor and rack space needed to house them and physical equipment management requirements.
Virtualization, then, holds a number of potential capital and operational cost-saving benefits. But it raises a few questions, too. For example, a mature virtual infrastructure will cross many traditionally separate internal groups of employees, such as those responsible for servers, networks, storage and security. So a virtualization project is likely to have an impact on organizational structure and responsibilities. Therefore, getting executive commitment and support, particularly from chief financial executives able to see the big-picture savings potential of going virtual, is critical to organization-wide buy-in and cooperation.
There are basic technical and logistical questions to consider when considering a virtualization plan:
• How do you calculate the appropriate ratio to use when consolidating physical servers and other devices into software-based virtual machines (VMs)?
• What is the impact of VMs on current disaster recovery and high-availability plans?
• Do security requirements shift in a virtual environment?
• How might software licensing/pricing models, power requirements and patch management processes need to change?
Preparing a well-documented business case based on an assessment of the current environment will help answer some of these questions. In your assessment and business case, you’ll need to calculate the physical host metrics needed – such as processor type and speed; RAM amount and utilization; network interface speeds and quantities; disk resources and other metrics.
The business case should also delineate the problems you expect virtualization to solve. For example, are you currently experiencing low utilization rates on servers while growing numbers of servers are becoming unwieldy to house and manage? Going through the exercise of calculating the hard benefits will help you answer some of these questions and drive acceptance and adoption of the virtualization project throughout the organization.
Let’s also take a look at each of these virtualization questions in a bit more detail.
Determining Consolidation Ratios
A primary benefit of virtualization is optimizing the use of physical equipment, such as servers, rather than letting them run underutilized much of the time and wasting money. However, it’s best to avoid extremely high consolidation ratios that push physical hosts to near 100% utilization. Rather, data center managers should leave some wiggle room for moving VMs around in the event that there is a planned or unplanned outage. About 65% to 75% utilization is generally a good average to shoot for to make virtualization worthwhile but safe.
Something to consider when calculating maximum consolidation ratios is moving off of small-form-factor blade servers and onto larger chassis, the enclosures for the server blades, to allow for greater consolidation density. An approximate maximum consolidation ratio of a 1U blade server is 6:1, for example, where a maximum consolidation ratio of a 4U chassis is 30:1. So consolidation ratios might be bigger, and savings greater, with larger chassis.
Note, too, that the deployment of emerging applications, such as real-time presence and video streaming services, might require greater network capacity for shuttling communications quickly among VMs. These applications might also lower the consolidation ratio that can be achieved compared to an existing data center environment without these applications in operation.
High-Availability Implications of VMs
The wiggle room afforded by retaining a spare 25% to 35% capacity helps address failover and disaster recovery when determining the appropriate number of VMs to consolidate onto a single physical host. Fortunately, the failure of a given VM on a physical host will not affect the operation of another VM on that host, because each VM has its own isolated set of compute, memory and power resources. So a VM could fail over to another VM on the same physical host or, in the case below, to a VM on a separate physical server.
Physical Redundancy
Virtualization, again, reduces the number of physical hosts in operation by consolidating multiple VMs onto a single physical machine. The greater the consolidation ratio, then, the greater the impact will be if a single physical machine should go offline.
For example, if a single server in the past hosted a single application and served a handful of users, only that one application and those users would be affected if the server went offline. If you put 10 applications and support for hundreds of users on a single physical device, however, and the host becomes unavailable, the failure affects many more applications and users, and has greater impact throughout the organization.
You need to have a disaster recovery plan in place to address that issue. For example, a given physical host might be configured to fail over to one or more other physical hosts. Traditional high-availability configurations often require a 1:1 ratio of primary device to backup device. A virtualized environment, however, could be configured such that a physical server can fail over to one of a set of backup servers, determined, in part, by the load balancing equipment that front ends the physical servers. This allows, then, for a many-to-one backup-to-primary configuration ratio, which increases service availability .
Security in a Virtual Environment
Moving to a virtualized data center means moving to two-tier security. The physical hosts must be secured, as always, and, now, so do the VMs. In unvirtualized data centers, most security is provided by special-purpose physical appliances that protect static resources, such as servers with unchanging IP addresses. In a virtual environment, however, enterprise security teams will need to deploy security tools that account for devices that might change addresses frequently in response to changing workloads.

cont....................