When it comes to server virtualization and cloud computing, network performance management may be an Achilles’ heel.
There is no question that server virtualization can help organizations reduce costs while streamlining operations. I constantly hear stories about server provisioning going from weeks to hours now that it is as simple as spinning up a new VM. In theory, this will get even easier as IT develops applications on top of cloud platforms like Azure, AWS, and Google App Engine while leaving all of the infrastructure gorp for others to deal with.
Unfortunately, there is a catch to this Polyanna tale. Server virtualization drives higher utilization of servers, more network I/O, and can distribute workloads that need to communicate. What’s more, VMs are mobile by nature. All of these things — higher density, more I/O, server-to-server traffic, and VM mobility make performance tuning an absolute bear. One VP of network engineering told me, “we are tuning the network all the time,” while another stated, “we did a deep dive on network flows and traffic analysis to understand how server virtualization was impacting network performance. We knew things were running slow but it took a lot of work to figure out why.”
Best Microsoft MCTS Training, Microsoft MCITP Training at certkingdom.com
This is a big problem. According to a recent ESG survey, 25% of organizations say that performance management is one of their biggest server virtualization challenges. Hybrid clouds? Burstable applications? Unless we figure out basic performance management in a virtual server environment, the cloud will remain a vision.
How do we address this problem? First, we need visibility up and down the stack — in real-time. Second, we need a better baseline understanding of network behavior. Note that I didn’t say “normal” behavior as nothing will be normal when infrastructure is shared across diverse workloads. We will need to extract operations data from virtualization and cloud platforms (like vCenter or Eucalyptus) and we will also need to store more packet capture data and marry network performance management with big data analytics. Finally, we will need standard ways to share and exchange network behavior data with cloud providers and ISPs.
Some of this work is in progress already — companies like NetScout, Opnet, and SolarWinds recognize IT requirements and are moving quickly to capitalize on the market opportunity. Riverbed may also have an interesting play here with a combination of Cascade and Wireshark. Vendors that are making progress understand that network dynamics are changing rapidly. Keeping up with these changes means capturing and analyzing data to a far greater degree than past LAN/WAN performance management efforts.
When it comes to cloud computing, everyone talks about security as being the biggest impediment. That may be true, but network performance management must become more virtualization aware and cloud-ready before the cloud computing show plays on Broadway. This IT challenge creates a huge market opportunity.