Why Network Managers should care about VMs

The last couple of weeks have produced a number of announcements indicating how the worlds of networking and virtual servers are coming closer and closer together, and how network management is adapting to the new challenges this creates. A quick note on the challenges: First, there is a whole mini-network inside each VM server, complete with a virtual switch and virtual NICs, and traffic between VMs that is happening without your ability to see it (hence, how will you manage it?). Second, “virtual sprawl” is happening, where the number of server images is growing much faster than they ever could in the past, meaning more end-points to watch when monitoring and troubleshooting performance. Third, VMs can be moved dynamically, from server to server, or even data center to data center, and you need to recognize that change when it happens so you can account for the consequential impacts on performance and capacity. Two weeks ago, EMC announced their Smarts Server Manager product, which finds and maps virtual servers and cluster, allowing the traditionally network-oriented EMC Smarts root cause analysis features to be extended to cover these increasingly popular elements. Last week, Cisco made their formal move into blade computing with their UCS launch, stamping “virtualization” on it in every way they could, and touting the importance of tight integration between virtual systems and networking technologies. Management was an essential element of this launch, although just how tightly their solution integrates management of the computing elements within UCS with the networking in and around it is still a little fuzzy. And today, Network Instruments announced a new version of their packet-based performance monitoring solution that can be deployed within VM environments. Though not the first application of packet analysis to vSwitch traffic (WildPackets launched OmniVirtual in May 2007), Network Instruments is including an innovative “Virtual TAP” feature which can aggregate all vSwitch traffic and ship it out a physical NIC, eliminating the blind spot and allowing security & compliance monitoring as well as full packet-stream storage for forensic performance analysis. Other vendors have added support for monitoring virtual servers into traditional network monitoring tools (you can find various levels of VM monitoring within offerings by CA, HP, Nimsoft, Solarwinds, Science Logic, Zyrion, and others), but this flurry of activity is a new step towards critical mass. Most support to date has been based on integral monitoring of VMs as "just another flavor of servers." Both the EMC and Network Instruments introductions represent the adaptation of higher value features within the network management arsenal, applying them to better understand and control VM ecosystems. With virtualized systems reaching mainstream adoption levels it is high time for networking teams to get serious about building an understanding of how their presence, which is certain to grow before it diminishes, affects their tools, practices, and policies. I’m interested in hearing what networking professionals are doing to deal with VMs. What are you doing to improve visibility and control of VMs? Are you monitoring them separately, or are you looking for integrated tools and collaborative practices? What has worked so far, and what hasn’t? Drop me a line with your thoughts and experiences, and I’ll share the results in future posts.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.

Copyright © 2009 IDG Communications, Inc.

SD-WAN buyers guide: Key questions to ask vendors (and yourself)