On a regular basis, when talking to practitioners, I hear complaints that automated orchestration functions (i.e., vMotion, high availability) within virtualized environments are being held hostage by networking configuration changes needed to maintain or restore full operations. (I’ve also heard that similar friction can occur around data and storage but this article is about virtualization).
Last month, EMA published results of a research project that looked directly into current practices for deploying virtual machines (VMs) and the networking related tasks that inevitably accompany them. The study was commissioned by Infoblox, and we were seeking a better understanding of whether or not the efforts to deploy VMs were being helped or hindered by network-side tasks.
VM deployment is a group effort
The first thing that we found was that VM deployment is clearly a group effort. It is not the sole domain of the server virtualization team, but depends on many other teams, as well. While 61 percent of respondents overall cited data center and systems administration as having the largest role outside of the server virtualization team, roughly 40 percent called out other groups such as network operations, application support, and network/application security as significant players in the process.
One of the more surprising findings was the volume of VM deployments that were taking place. About a third of respondents were deploying up to 200 VMs per month, another third between 200 and 500, and another third between 500 and 1,000 VMs per month. Higher rates roughly correlated with larger organization size, but not exclusively. As VM deployment rates grow, this will increase the need for higher levels of efficiency in order to deploy, manage and troubleshoot them when things go awry.
A quick story to illustrate this point: The Infoblox CIO reviewed the survey with his own server virtualization team and the CIO was astonished to learn just how many VMs per month were being provisioned at his own company. It is reasonable to speculate that the same gap in perception versus reality could and probably does exist in other companies.
So where were the long poles and the short poles in the VM deployment tent? Turns out that it depends on your point of view. IT execs and sysadmins believe that VM provisioning itself was the longest step in the process, however networking pros felt that setting up the VLANs took longer than VM provisioning.
Best practices are still evolving
Perceptions regarding most common root causes of VM deployment failures also depended on point of view. IT execs felt that VM configuration errors were the most common root cause, whereas networking staff tended to point at storage, and sysadmins pointed at the network. This lack of consistency indicates that there is little true systemic analysis being done and that true best practices are still evolving.
Participants were then asked questions about the level of collaboration that existed when work processes crossed team barriers. Responses indicated that for the most part, processes had been formalized (versus being ad hoc) but were largely manual (versus automated or orchestrated). It’s good to see that IT groups have indeed established some degree of cross-team operations protocols, however, the lack of automation will impair efficiency and become an increasingly significant barrier to deployments at scale.
Making things better
So how could things be made better? Cross training is often a means for achieving better clarity, however 70 percent of respondents indicated they had yet to complete such efforts, and nearly 20 percent had no such training plans in store. That said, a third of respondents said training was a high priority, even though the majority indicated they lacked the tools necessary to share information and collaborate proactively.
Clearly there are opportunities for improvement here. Many of these challenges can be met with network change and configuration and run book automation tools. A bit further off in the future, software defined networking (SDN) could also help. But in the immediate here and now, the quick answer lies in picking up the phone or peeking over the cubicle wall and opening a direct dialogue between the networking and virtualization teams.
Until technology comes to save the day through automation, only direct human engagement will remove the immediate barriers that hold back the true potential of server virtualization technologies.
Jim Frey is a research director at Enterprise Management Associates. Jim has over 24 years of experience in the computing industry developing, deploying, managing, and marketing software and hardware products, with the last 18 of those years spent in network management, straddling both enterprise and service provider sectors. At Enterprise Management Associates, Jim is responsible for the Network Management practice area. Prior to joining EMA, Jim spent six years with NetScout Systems as vice president of Marketing.