Fixing Your Network’s Five Worst Bottlenecks
Networking Pipeline
November 14, 2005
Got a bogged-down, sluggish network? The problem likely isn’t that you’ve outgrown your infrastructure — you have some serious bottlenecks. Here’s how to fix your network’s five biggest bottlenecks.
By Matthew Friedman Networking Pipeline
It’s so plaintive that it can be heartbreaking: “Why is the network so sloooooowwww?” plead users in just about every organization in North America. The inability of a network to keep up with the expectations and demands of its users seems, at times, to be the defining characteristic of networks.
“Everyone, at some point, complains that the network is too slow,” Info-Tech Research analyst Carmi Levy says. “Very often, organizations think that they’ve outgrown their network, without considering that the real problem is that it’s a bottleneck,” that is causing the problem.
Indeed, network bottlenecks are so common that they’re almost a fact of life. And because of the nature of the problem, many organizations simply fail to address the problem, choosing instead to simply throw money at upgrades and hope for the best. “It’s not usually a big-bang problem,” Levy says. “Organizations aren’t experiencing massive bottlenecks that compromise their ability to do business.
But they are experiencing slowdowns that can often stimulate unnecessary spending, and that can affect the bottom line. “Instead of dealing with the problem, companies often just go out and buy new stuff. That’s spending inefficiently because they’re not doing a cause analysis.”
And without taking a close look of the fundamental causes, the bottlenecks are unlikely to go away, no matter how much money you throw at them. The key to really solving the issue of course, is to know where to look and, although there is a staggering variety of them, some kinds of bottlenecks are both more common and more pressing than others.
1. Poorly Tuned Servers: While it’s easy to think of the network solely in terms of the plumbing, it’s important to remember that its purpose is to move data around, and where that data reside can be a big source of network slowdowns. “The problem is that servers are being asked to do more than ever before,” Levy says. “They’re handling all kinds of data and applications and multiple layers of traffic, and they’re expected to do it all well.”
The key to server performance is to understand just what it is that you expect each server to do, and set them up accordingly. Despite the promises of out-of-the-box performance, the real world is never quite that simple. “You have to configure for optimum performance for every job that every server is supposed to do,” Levy says. “It’s can be time consuming, particularly is you’re running a whole lot of servers, but the network performance gains are worth the effort.”
2. Constellations of Appliances: Every network function, it seems, has been condensed into an appliance, and that can be a problem. “There are more devices on networks today than ever,” Levy says. “Every time we need some kind of new functionality, we toss new stuff on the pile.”
Security appliances, firewalls, Blackberry servers, the Google search appliance: Each appliance that does what used to be managed in software adds processing time to the network signal. “Vendors are selling a solution in a box,” Levy says. “That can screw up your network balance. It’s worse that it was, and it’s only going to get worse.”
The solution is to decide what you can do without and what you can leave to servers. Appliances can be very useful and convenient but, Levy says, when the short-term fix contradicts long-terms goals like network performance, you have a problem. “Slipping a best-of breed solution to address tactical issues only make sense if it doesn’t run counter to long-term needs,” he says. “That’s why you have to stop and think ‘Do I need this device?’”
3. Improper Segmentation: Network performance wasn’t a major issue when you had relatively few users and devices using relatively few resources. But the increasing demand placed on networks has made the plug-it-in-anywhere approach a recipe for bottleneck frustration.
“This is an increasing issue, especially as we’ve become more security conscious and have to set up secure zones,” Levy says. “But a lot of networks are built without a roadmap, so they’re either not properly segmented, or not segmented at all.”
The idea, of course, is that network performance will only suffer when everyone’s traffic is running over the whole network. If the graphics department is doing some heavy rendering, there’s no reason for all of that traffic to clog up the accounting department. Better to give the arty guys their own little patch of Ethernet than have them take over the whole company.
“It’s important to invest in areas that need performance and segmenting them away from areas that don’t,” Levy says. “This is about planning and executing your network as efficiently as possible. You don’t need to build a superhighway to get to the grocery store.”
4. Misbehaving applications: Even if the plumbing is optimized to within an inch of its life, there’s always the danger that your network applications are bigger, more bloated, and more inefficient than they need to be. Anything that runs on the network affects its efficiency, and anything that uses the network inefficiently will have serious performance consequences.
“The sad truth is that a lot of application developers develop applications without understanding, or sometimes not caring, what their impact will be on the network,” Levy says. “Those applications steamroll everything else. A sales force automation application that synchronizes a database by moving the whole database over the wire is going to cause problems. There are more efficient ways to do these things, and applications that don’t do things efficiently are going to cause huge bottlenecks.”
It’s important, then, to try before you buy. Levy says that the promises and benefits of a new application have to be weighed against its network impact, and the only way to measure that is to test it out. “This sis something you need to consider before deploying an application,” he says. “Do you have a testing network where you can analyze network functionality before deployment? You should.”
5. Bad Security: Apart from the very real dangers to your corporate data itself, security is very much a performance issue. “Have you ever tried to run anything on a network where the PCs are running as a zombie net?” Levy asks. “It’s a lesson in frustration.”
With keyloggers reporting home, adware pulling in graphics and data from the Internet and spam clogging mail servers — none of which add anything to your business — you can have a very big performance problem. “Insufficient security does lead to the inefficient use of business and network resources,” Levy says. “It bleeds them dry.”
Whether any of these issues are at the root of an organizations network bottleneck woes is hard to say. However, Levy says that the only way to start to answer the question is to start looking at how the network actually works, and whether there are problems that can be fixed through good management practices rather than technology.
The biggest bottleneck is the lack of proactive network administration. “Before you throw money at anything, you need to do a protocol analysis, and know what’s happening on the network that you have,” Levy says. “Connect sniffers, understand where the traffic is coming from. You can’t manage your network unless you know how its performing, and that should be how you guide your efforts. If that analysis says ‘yeah, we’re at capacity,’ then you can go shopping for new hardware.”