In the days of physical servers, it was in some ways much easier to manage the relationship between an application, the servers upon which the application ran, and the physical connections to those servers. But with the use of server virtualization, when a single physical server runs multiple applications, and when at any given time an application could be running on any one of a number of servers, how does one manage these virtual network relationships? Data center network managers now need a way to "reconnect" physical resources and virtual workloads in order to understand how they are being utilized and how to manage them.
The need, from a management and monitoring perspective, to connect physical resources with virtual workloads drives an entire portion of the virtualization industry. A number of companies are working to provide products that offer an "end-to-end view" of physical resources and how they are being utilized by the various virtual workloads at any given time. Because this utopian end state has not yet been achieved, it's important to consider what steps the IT organization -- not just the data center networking team -- can take in the meantime.
Documentation is critical
First, extensive documentation is a critical component. Good, thorough and understandable documentation is important to any information technology project or implementation; it's particularly important to server virtualization projects and the impact of those projects on the data center network. This documentation needs to include a detailed summary of the network configurations in support of the virtualization project, such as where link aggregation -- like 802.3ad/LACP or any proprietary equivalents -- was used. Information on VLANs and 802.1q VLAN trunks should also be included.
In addition, data center network managers should take advantage of any other network protocol support provided by the virtualization solution, such as VMware ESX's support for Cisco Discovery Protocol (CDP). CDP support in VMware ESX 3.5 and later provides a way for data center network managers to map the relationship between the host server's NICs and the switch ports into which they are connected.
Limit live migration
Second, the server administrators may need to work with network managers and other teams to regulate more closely the use of technologies that dynamically move VMs between different physical servers. Live migration -- the ability to move a running VM from one physical host to another -- is generally considered helpful, and some vendors have expanded upon live migration to include the ability to automatically shuffle VMs around in response to resource demands.
Until such time as the end-to-end view becomes readily accessible, however, server administrators may need to be more conservative in their use of dynamic technologies. While conservative use can limit the flexibility of the virtualized environment, it helps to simplify the mapping of applications to physical resources and makes troubleshooting as well as performance management a bit easier.Investigate management tools
Though it's true that some network visibility can be lost by the introduction of server virtualization, it's also true that a number of products have appeared that can help restore this network visibility. The ability to view network traffic patterns and flows can be a key part of understanding the relationships between various applications and the underlying physical resources. Organizations that rely heavily upon such information should investigate the use of these various third-party products that provide information and insight into virtual network traffic. Organizations should also take the time to investigate applications that claim to provide the end-to-end view of applications, virtual machines and physical resources. This type of application is still developing and evolving, but there are some good implementations already on the market. These applications gather information from a variety of sources, both physical and virtual, and correlate that information to help organizations understand which applications/VMs are using which physical resources. Based on the correlated view, data center managers can adjust the configuration of the network to accommodate traffic flows and connection patterns more properly.
As server virtualization continues to mature, more applications will be released that do an even better job of providing data center managers with the information they need to view, understand and manage network traffic patterns and how those patterns reflect applications running inside VMs on various physical hosts. Greater integration between the server virtualization vendors and the data networking vendors will also provide even more ways to map VMs to physical resources, thus helping to address this key business need.
About the author:
Scott Lowe is a technical lead for virtualization at ePlus Technology Inc. During his 15 years of experience, he has worked with a broad range of products, specializing in enterprise technologies such as storage area networks, server virtualization, directory services and interoperability. In addition to his work for SearchNetworking.com, Scott is a contributor to SearchVMware.com and SearchServerVirtualization.com. Scott also maintains a blog about virtualization, storage and servers.