The Impact of Virtualization on Wide Area Networks—Part 1
While server virtualization continues to redefine how (and where) IT and data center teams deploy computing resources, a gap in organizational boundaries still exists in most enterprise and public sector IT organizations that results in “the left hand not knowing what the right hand is doing”. In other words: server and storage infrastructure teams are saving capex and time through virtualization, but the network team probably hasn’t modeled the impact of those changes into their wide area network (WAN) budgets or projects. Analysts writing for Network World offered similar feedback in a recent article series.
For the last few years, IT teams have started consolidating branch IT resources into the data center—Exchange, business application and file servers, as well as related storage. This architectural trend in increasingly mitigated by implementing a WAN optimization solution (like Cisco WAAS) to offset the increased bandwidth and WAN latency added with centralizing resources and requiring remote users to access the data center.
But server virtualization add a new wrinkle to this trend, as IT projects can be easily combined and budgeted at the server team-level. As we know, most branch servers are operating at less than 15-20% CPU capacity, and even most [non-virtualized] data center servers run at less than 30% capacity. So it’s an easy and safe IT project for server teams to consolidate dozens of branch-based Windows and other utility servers onto a few data center-based multi-core servers with VMs turned on.
And this challenge will only get worse as virtual server deployments and resulting strategic architectures evolve to include VM portability and public/private cloud computing. Imagine that a Singapore-based sales manager is accessing a CRM application which was once local, but is now based in the Hong Kong data center. But it’s end of quarter and the peak load on the Hong Kong CRM application is over the specified capacity limit, so the transaction (running on a virtual machine) is transferred via VMotion to the data center in San Francisco, completed and sent back to the Singapore sales office.
As IT teams continue to adopt and increasingly leverage virtualization to improve asset utilization and agility, it’s going to be key that organizations themselves virtualize and become more agile, sharing their own “IT project transactions” transparently so a decision from one team (i.e. servers) doesn’t severely impact the “up time” or “TCO” of another team (i.e. networks). The CIO of a major US bank discussed this issue in CIO Magazine and how they addressed these challenges.
Do you have experience or results to share on how IT teams can adjust or “virtualize” their own organizations to best get ready for the next stages of IT virtualization? Or prepare across different IT teams and assets to ensure that everyone is “virtualization-ready”?
Posted by Mark Weiner at 10:41AM PST
Stefan Apr 28, 2009
Very good subject! I am testing right now fat clients coming across WAN after SMB/CIFS resources in the Data Center, using WAAS, vs. VDI deployment in the Data Center, where the resources are, both over site-to-site VPN. Attempting to derive cost vis-a-vis performance vis-a-vis scalability (WAAS size vs. server size to sustain VDIs) info.