Solution provider takeaway: This tip outlines the top three things a networking solution provider must keep in mind when it comes to virtualization networking. It's important to keep an eye on network performance because consolidating servers in the data center can affect the network in many ways.
Somehow it seems counterintuitive to worry about network performance when getting ready to consolidate servers in a data center. After all, if you're taking out server hardware, logic suggests that demands on the network would be reduced.
Network specialists know that this thinking is seriously flawed. In fact, intelligent network design is a fundamental part of successful server consolidation projects.
"One of the hamstrings of virtualization so far has been network [input/output] and data center I/O in general. ... These things do have to be taken into consideration," says Mauro Lollo, co-founder and chief technology officer of Unis Lumin, a network integration company in Oakville, Ont.
"You can get into an ugly place through poorly planned projects," says Tim LaFazia, president and CEO of Infrastructure Development Corp., a network integrator in San Diego. "A lot of it has to do with optimizing bandwidth capacity."
In that vein, here are three basic things that network integrators can do to ensure that their customers' networks are properly considered when planning data center virtualization projects.
#1: Make sure there are enough ports to handle VMware server traffic
Pete Thompsen, senior engineer with solution provider CT Networks in Hauppague, N.Y., says many businesses going virtual are thinking about using Gigabit Ethernet networks to support them. It is possible to support virtualized traffic with 10/100 Ethernet switch ports if the network is properly designed, however, especially if the business is going completely virtual and using thin clients, Thompsen says.
Regardless of the underlying network technology, network integrators say it makes sense to perform a full networking equipment audit of switches, routers and edge devices, taking into consideration all of theavailable network ports for VMware servers as well as the scalability of these devices.
In particular, Thompsen notes networks where core networking devices are daisy-chained together haphazardly might pose a challenge when virtualizing. For the best performance within a virtualized server environment, edge devices should connect directly into core networking equipment via a star network topology, he suggests.
#2: Plan virtual server distribution so it doesn't overly tax physical resources
LaFazia says some businesses considering virtualization projects look to their original existing network architecture blueprints when they get started only to find there is more than meets the idea, or that the design has changed over time and wasn't properly documented.
"Before anyone gets involved, they first really need to have an understanding of what they have. They need a true audit," he says.
That's because you need to take into account both the components of the physical servers and the workload types of the virtual machines that they will host when considering the implications for network performance. Having two high-bandwidth virtual images running on the same physical server, as an example, is not a great idea. Unintended network latency can be introduced when capacity is not distributed equitably, LaFazia says.
Network integrators can also help businesses understand which server assets they no longer need, rather than simply migrating everything to the virtual world, LaFazia adds. "Easily 10 to 13% of the server assets that they are migrating have lost their relevance to the business," he says. "But they don't know this. Doing this will help clean up the scope of the project. The core message is clean house first."
#3: Don't forget security
Ensuring that the proper level of security is maintained when virtual machines are moved from server to server is also critical, says Lollo.
That means firewall settings for each network device need to be considered carefully for compliance and audit purposes. Organizations need to ensure that as virtual machines move from physical server to physical server, the same access controls are applied.
"The drag-and-drop approach of virtualization means you could be running in one data center one minute, then another the next," Lollo says.
Indeed, that's the big promise of business continuity and disaster recovery solutions that use virtualized servers as their foundation.
In addition to implementing firewalls that can migrate, Lollo says adequate security and fault tolerance measures must also be built into the physical network topology.
"You can't just assume that plugging into another firewall will be adequate. You really have to sit down and think about the balance of the environment," he says.
This was first published in October 2010