Voice, video and data are beginning to converge on the same network. How does that change the need for network performance-monitoring tools?
Voice is not a very high bandwidth application. With the addition of voice, network managers are trying to contend with an application that is suddenly very sensitive to latency, jitter and packet loss. If it takes a little while to get e-mail, that is not a big problem. The typical data application is not time sensitive. With a phone, people expect to pick it up and get a dial tone. They are not getting a lot of new functionality with a voice over Internet Protocol phone, so they are not willing to sacrifice any quality. There has been a whole paradigm change in terms of what these applications are sensitive to, so there needs to be a completely different approach to network management. What are the most important performance criteria in that environment?
Measuring voice quality is a tricky thing to do, since it is very subjective. You usually string together quality of service (QoS) and mean opinion score (MOS) [which is used to measure voice quality]. Right now there is a benchmark at 4.5 MOS out of 5, so you set your managements tools to watch for that.
Also, network performance needs to be tied to business processes. For example, if a router goes down and a database server slows down [simultaneously], it seems to make sense to fix the router, since it is down, but perhaps the router is redundant. Meanwhile, the database server supports mission-critical transactions, so the number of transactions that can be processed drops. So network management needs to be tied to business function. What technologies on the horizon should people be keeping an eye on?
In general, network managers should keep an eye on those technologies that are targeted at managing quality of service. A lot of vendors are trying to go that way right now, but it still largely remains a black art.
FOR MORE INFORMATION:
Browse our Topics on quality of service
Read our free white paper on Migrating corporate voice traffic to the data network: Strategies, risks and rewardsDoes that change how network managers should think about their networks?
Network managers need to move from being passive and reactive to being active and proactive. You have to combine network management and quality of service management. Often management tools lag behind new applications by 12 to 18 months, but with IP telephony, you can't afford to have that time lag. It's fine to have the system falter in the first few weeks, because you can switch back over to the old phone system, but if you have a problem nine or 12 months down the line, usually that old system has been replaced, so it is a serious problem. So you have to have the management tools in place. How do you balance the need for consistent quality with tight budgets?
We are seeing baby steps with implementations, smaller departmental or regional rollouts. Companies are seeing if they can get this to work and how much it costs. Companies are also starting to see that they are not going to save a lot of money with VoIP. They are moving beyond the cost story to look at how this helps produce better service for their customers. ROI is hard to measure. What does it mean that a sales rep can spend four more hours a week making calls? Based on the percentage of calls that become sales, there is extra revenue being generated there. The CIO should be able to do the math. How are vendors adapting to this?
With IP telephony, you need active monitoring. You need to be able to emulate voice calls on the network to establish benchmarks, and you have to constantly monitor the network. And you need to do all of that without stressing the network. Vendors are taking a number of different approaches. Some use agents, some the client-server model, and some use distributed, peer-to-peer environments.