Network Instruments started conducting “State of the Network” surveys about three years ago, asking questions in two main categories: network-troubleshooting concerns and emerging-technology adoption. This year’s study, surveying 592 respondents (IT managers, network engineers, CIOs), shows some interesting trends in the networking arena.
“What’s the primary concern when a network engineer begins troubleshooting?” asked Charles Thompson, manager of systems engineering. “According to our survey, 75 percent of engineers say the biggest concern is simply identifying the source of the problem. That’s a 25 percent increase from 2007. I was very surprised by that. And 71 percent said they spend at least 17 hours per month trying to identify the sources of problems. That’s before any troubleshooting begins! These days, a lot of network engineers have the wherewithal to troubleshoot problems, once they know where to look. But the problem is that there are so many potential sources: routers, switches, firewalls, NAC, application servers, database servers, load-balancers—aaaahh!”
Thompson then dug a little deeper. “Often, when you troubleshoot problems, you put an analyzer in place, get on the phone with whoever has the problem, and see if you can re-create the problem. Just like when you take a car to the mechanic: He has to hear the knocking you’re hearing. Unfortunately, problems are usually sporadic. According to the survey, 32 percent of engineers spent at least 400 hours per year trying to replicate problems! That translates to 20 percent of that engineer’s time just replicating problems.”
These results play nicely into the hands of an emerging solution in the networking realm: retrospective analysis. This kind of “TiVo for the network” solution, always catching data, throws that whole time-consuming replication problem out the window. “You can just look at the TiVo,” Thompson said, “and ask for the data that happened this morning at a certain time so that you can start troubleshooting. Retrospective analysis will become the ubiquitous, de facto standard in the industry.”
Another focus of the study was emerging technologies in the industry today. The strong rate of VoIP implementations increased 5 percent in 2008, with 66 percent of organizations already having implemented or looking into implementing VoIP within the next 12 months. Quality of service and the impact of VoIP on other applications proved to be the largest VoIP concerns for network professionals. This was a shift from 2007, when reliability was the greatest concern. The most notable shift occurred with the network professional’s confidence in their VoIP systems. Only 13 percent in 2007 were completely confident in their system, compared to 25 percent in 2008.
Thompson said, “To me, a big surprise was how high the adoption rate is for 2008. But of the current VoIP adoptees, who has visibility into VoIP traffic? Well, 38 percent said they had visibility, but 62 percent said they had no visibility. How does that play into network visibility? VOIP is a real-time application—no buffering, no cushioning for time. I need to deliver that in as timely manner as possible. The unidirectional delay budget is about 150 milliseconds. If it takes me longer to get the packets of my voice talking to you, we end up with audio clipping, and we start to talk over each other. The real-time delivery of the app and the sensitivity are big concerns. The other big concern is the user tolerance of poor voice quality. We’ve all had delays pulling files off the server or accessing IM or email, but we’re not as tolerant of not getting a dial tone for a phone call.”
What about Multiprotocol Label Switching (MPLS)? Migration to MPLS networks on a global basis appears to be steady, with most organizations still in the early stages of adoption. Thirty-five percent of respondents will have migrated to MPLS networks in the next year, while 55 percent don’t intend to migrate to MPLS.
Thompson said, “I thought that 35 percent figure was kind of low! I mean, the cost is low. What’s keeping people away from MPLS? They tend to say, ‘We don’t fully grasp MPLS, and we don’t have time for the training.’ Also, a lot of organizations don’t understand the need for MPLS. The point is to mesh the network, to create point-to-point connections between remote facilities so you don’t have that hub-and-spoke topology. But that mesh notion doesn’t compute a lot of the time.”
The study also looked at and 10 Gigabit Ethernet (10 GbE) adoption. Twenty-four percent of organizations will implement 10GbE networks within the next 12 months, while 71 percent indicated they had no plans for deployment.
“I was surprised by the relatively high percentage of 10GbE adoption rates,” Thompson said. “Network engineers are trying to migrate away from two switches designed to communicate over multi-connection 1Gb link (so they would bond together two or three 1Gb pipes into an Etherchannel) and upgrade that connection. So it’s really the connections between switches that 10GbE is being rolled out to. I’m not seeing a lot of 10GbE to the server or desktop. It’s mostly switch-to-switch communication.