The thought of network capacity planning usually brings up ideas of new systems or budgeting processes. In a highly...
virtualized data center, however, understanding the current network load is one of the central pillars for optimizing the virtual environment.
Network capacity planning has traditionally been about estimating the growth of the data center network in order to budget for the number of network ports needed during the coming year. The data center team could easily identify the servers that generated the highest network traffic, and network administrators in many cases could simply throw more bandwidth at those with planned network upgrades. They would translate yearly capacity planning calculations into budget requests, and then set aside the formulas until the next budget cycle.
Capacity planning for a network that supports virtualization accomplishes the same task, but at a near real-time pace. Virtualized environments like VMware abstract the compute, memory, storage and network resources of the data center and share them among multiple servers and applications. The environment is designed to treat all of these resources as a whole to be divvied up among the virtual machines to optimize utilization. Without a clear understanding of the utilization of the network, it is possible for a server administrator to group multiple virtual machines that are not CPU-intensive on the same physical server. But those virtual machines might be network-intensive, risking a saturation of network I/O for that server. Doing proactive monitoring of the physical and virtual networks, and sharing that information with the rest of the virtualization team, is the only way to ensure that the network does not become the bottleneck.
New capacity planning tools for the virtual world
The industry needs a new generation of network monitoring and capacity tools to link the physical and virtual environments in the data center. Virtual machines are highly mobile in the virtual environment and can literally spin up anywhere on the data center. To analyze the real-time impact on the network and begin to plan for future growth, network monitoring and capacity planning tools have to collect data from the physical hardware and communicate with the virtualized environment management tools.
“A virtual environment needs more than SNMP for monitoring and capacity planning,” said Kenneth Mills, principal product marketing manager for service assurance at CA Technologies. According to Mills, integrating capacity information from the hypervisor, storage and networks on a single pane of glass is critical to managing all of the resources in the virtual environment quickly. This consolidated view allows systems administrators to make informed decisions on the best location to move or spin up virtual machines.
Network capacity planning nets real-world results
“Capacity planning and management is a must have for virtualized environments,” said Rick Parker, IT director for Fetch Technologies. The California-based company collects unstructured data from the Web and delivers it in a structured format for its clients, making its virtualized data centers mission-critical to the organization. By collecting nearly 3000 monitoring metrics from the data center network and virtualized environment, down to the physical and virtual port level, Parker can detect capacity trends that give him advanced warning of additional capacity needs. “If I see a steady 10% rise in traffic, I can map out how soon I will need to bring additional resources online,” said Parker.
Parker uses a mix of monitoring tools, including Solarwinds’ Virtualization Manager and ipMonitor, Alertra.com for external monitoring, as well as the monitoring and reporting tools within VMware’s vCenter, to keep an eye on his virtualized environment and his data center network.
Unlike the single pane of glass view offered by a product like CA Technologies’ NetQoS Performance Center, Parker's tool set gathers and analyzes metrics from the respective environments to give a whole view of the data center. The reports and alerts generated from these utilities enables Parker to single-handedly administer a Brocade VDX fabric, running 350 virtual machines in multiple data centers. Parker receives email alerts about problems, as well as daily, weekly and monthly reports on the overall capacity of his environment.
This level of network capacity planning has allowed Fetch Technologies’ IT organization to avoid operating on a fixed IT budget. Instead, Parker relies on both capacity planning reports and line of business drivers to dictate hardware expenditures. “When sales bring on a new client data request, we can spin up the needed capacity before the ink is dry on the contract,” he said.