Rick Vanover, Contributor
Published: 29 Jul 2010
Administrators always search for incremental performance enhancements and operational efficiencies to fine tune a virtualized server infrastructure. As such, seemingly minor adjustments and tweaks can ensure that a complex virtual environment maintains a high level of performance for users and clients.
This article showcases five ways to optimize a virtualized server infrastructure's effectiveness. These configurations are applicable to all virtualized server environments. While some practices may seem small in scope, they can greatly affect your infrastructure performance.
1. Boost virtualized server infrastructure performance with file server consolidation
File servers can be a big pain point for administrators. Often, when file servers are consolidated from physical systems, they're simply converted and loaded onto a virtualized server infrastructure. As a result, administrators miss an opportunity to delete, consolidate and organize potentially unstructured data. So let's consider a slightly different approach to optimize file servers.
One option is to use a storage area network (SAN) to perform file server functions. Many storage systems can serve as a Common Internet File System (CIFS) endpoint on a network for an easy connection to an Active Directory domain. This method virtualizes all of your file servers.
Another consideration is a cloud-based virtual file server. With this approach, you should allocate approximately 10% of your file server storage to local machines and archive the other 90% in the cloud. From a functionality standpoint, a virtual file server presents the entire storage contents as if it were stored locally on your network. Furthermore, a virtual file server can connect to Active Directory and support full NT File System permissions for easy management, similar to physical servers.
The final option is to run a virtual machine (VM) as a file server. In this configuration, you should consolidate the computer names to reduce the number of file servers. Windows systems are capable of serving multiple computer names, or NetBIOS names, that users connect to through a short name (e.g., \\ServerName).
Many administrators prefer a large file server instead of administering multiple, smaller file servers. Windows Servers can serve additional NetBIOS names by adding the OptionalNames value in the HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\services\LanmanServer\Parameters area of the Windows registry.
2. Disband Microsoft clustering to improve infrastructure performance
It's common to convert a pair of servers running Microsoft Clustering Services (MSCS) into a VM. MSCS provides application-level failover within a VM that isn't possible with VMware High Availability and Fault Tolerance. But this arrangement may be redundant for many virtualized server infrastructures.
For VMware environments, MSCS is one of the few configurations that require a raw device mapping virtual disk in lieu of the popular Virtual Machine File System data store. This setup complicates storage provisioning, but in a minor fashion.
Disbanding MSCS clusters, however, has a hidden benefit: You can reduce your licensing costs. Your Windows systems are probably licensed at the Enterprise level or higher. I the MSCS cluster is moved to a VM that runs Windows Server Standard, you can lower your licensing renewal costs.
3. Improve OS management in your virtualized server infrastructure
Fine-tuning your virtualized server infrastructure with various management tools and processes can pay major dividends in terms of your infrastructure performance.
Having clean Group Policy configurations with limited local customizations is a best practice, for example. This makes for very quick server deployment through a VM template.
Enterprise system management software also helps virtualized server infrastructure deployments. Most virtualization administrators don't want to set up program installations as well as map network drives and localized configurations.
And while Active Directory's Group Policy is sufficient for most small environments, larger environments use Microsoft System Center, Altiris and Novell ZENWorks management utilities. These tools coordinate application delivery and manage interaction with guest VMs via policy.
4. Plan antivirus scans of your virtualized server infrastructure
There are several commonplace practices in nonvirtualized environments that snowball into major issues for consolidated virtual environments. To circumvent notorious antivirus scanning problems, for example, consider a policy that performs a full-system antivirus scan at 3 a.m. on Sundays for every server (assuming that's not a busy time for your organization in some other part of the world). This principle can also be applied to physical servers, because their processors are likely idle at this time.
Additionally, antivirus scans can raise concerns in SANs, because they place a tremendous amount of strain on local processor inventory. If you rearrange the pieces in a virtualized server infrastructure, you'll realize that the host processor inventory is severely affected when every guest simultaneously runs a full scan.
To fix this issue, stagger scans across a group of VMs. Schedule 15 VMs for a 3 a.m. scan, for example, another 15 VMs for 5 a.m., and so on. It's important to schedule these scans in such a way that technologies like VMware's Distributed Resource Scheduler can migrate workloads based on cluster compute capacity.
5. Improve infrastructure performance by forgoing P2V conversions for new builds
A physical-to-virtual (P2V) conversion is a very straightforward process, but a new build is always cleaner. In Windows systems, a P2V conversion leaves many residual configurations -- such as hardware drivers, unnecessary software and file system clutter -- that can adversely affect system performance.
To achieve the optimum performance from new systems, it's best to deploy a new VM and provision the required software, including the reinstallation of custom applications, transferring code and configuration information, and database restores on new database servers. Still, P2V conversions are very useful because many administrators don't have the time to perform a new build.
About the expert
Rick Vanover (email@example.com) has the vExpert, VCP, MCITP, MCTS and MCSA certifications. He is an IT Infrastructure Manager for Alliance Data in Columbus, Ohio and is an IT veteran that specializes in virtualization, server hardware, operating system support and technology management. Follow Rick on Twitter @RickVanover.