Dr. Cathy Fulton is CTO of NetQoS, a startup company that is building a new generation of network management tools. The trouble with these tools, historically, is that they have been dependent on agents to gather data, which are problematic to deploy, manage, and maintain across a network. In this interview, Fulton talks about how NetQoS has developed a new approach to agent deployment that will allow IT organizations to better manage any network in the world.
Q: What problems are you trying to solve?One of the problem areas that we see is there's just a ton of statistical data on the network that's available, especially application layer information. But there's no real good tool to manage this information or produce any sort of useful analysis or recommendations as to what you can do to correct a problem. What we add is that we provide some software which allows people to, for instance, optimize the configuration of the routers, or find out where the problems are on their network, and tune it so that they can optimize the return on investment they get from things like routers. We're really all about optimizing the resources that you have on your network.
Q: How is that different from people already selling tools in this space?Our customers, for instance, are typically Fortune 1000 types [who] have very large global WANs [wide area networks]. We help them find out what's going on their networks and how they can tune the network itself in order to improve performance. We have two products. One, Reporter Analyzer, is more strategic [and] takes information from a variety of sources across an enterprise network to find out, for example, if a legacy application is transferring data quickly but, as a result, is filling the pipe and starving off all the other business-critical applications. Our other tool, Super Agent, is directed towards quickly identifying where a problem is. The neat thing about Super Agent, our responsive time analyzer, is that you don't have to deploy agents across this global network in order to get response time information. You get it by placing our Super Agent just at the server farm, without deploying agents, and you see everything that the clients are doing across the network to the server farm and can get a breakdown, not only as to their response time and server throughput and loss performance, but also [as to the] volume of information.
Q: As an industry, where are we in terms of performance management on the Internet?The answer to your question, particularly regarding performance management, which is where our expertise is, is that it's really in its infancy. It's a very young, immature area. Everyone wants to get some automation, and I'd certainly expect us to have something to show you within a year or so regarding that. But the neat thing about performance management is because it is such a young area there's just a lot of growth and opportunity for any company, really, not just us.
Q: So is that something you are actually working on?Absolutely. But this is one of those questions where I don't really want to talk about that product yet because it's not in prototype yet. But that's absolutely where we're going. You can imagine we have this device sitting at the server farm that sees all the clients accessing the server farm and we have the metrics that the clients care about. So clearly, given that information and given configuration information about the network, you can tell where we're going.
Q: Do you plan to integrate your tools with other network management consoles?We just released 2.0. With this release, we do have a multi-unit architecture where you have individual collectors and our own console. We do send SNMP traps, for instance, to other platforms. Some of our customers are interested in HP OpenView's Integrated Service model of operations and kind of integrating it within that. But right now the extent of our integration is using SNMP.
Q: Why is it important for network management tools to examine the application layer?We always look at the application layer in making our decisions because different applications have different requirements as to what resources they need in order to operate properly and in order for the users to be happy. Some need large bandwidth, so you want to push it across fairly quickly. Others are sensitive to delay, others are sensitive to loss. You really need to look at the application layer in order to make intelligent decisions as to how you should tune your network.
Q: So what one thing would you like to see IT organizations do better?I would like people to think more about what they can do. If there's a problem on the network, you really need to know what the problem is to address it properly. It's just like an automobile traffic system. If you have a lot of congestion, you need to build new roads. If you have short periods of congestion, you can handle it with light synchronization. I'd like people to become more aware of the options they have in handling the problems on their networks, not to think just simply about increasing the bandwidth. There are things like service scheduling, dynamic routing, and packet shaping. I have a whole list of tools and approaches that I might use in order to get better utilization. But people are so busy fighting fires that they don't really have time to think more strategically, so they just react.