Bandwidth Management Techniques — Tips and Actions

Traffic shapers limit the speed and bandwidth available to certain data streams.
You can limit a particular type of application (e.g., specify that traffic from file­-sharing software can never exceed 200 Kbps), or you can specify how much bandwidth is available to each user. This technique is sometimes known as packet shaping or bandwidth limiting.
  • Limiting by type of application (or port number) is occasionally difficult because software can be designed to use random port numbers to communicate in order to evade traffic shapers.
  • Also, port 80 (the HTTP port for standard Web traffic) handles a wide variety of different data, and you don’t want to throttle it all indiscriminately.
  • Most traffic shapers can handle these routine evasive maneuvers by inspecting each packet of data carefully.
  • Investigate your packet­-shaping tool to see what kinds of distinctions it can make. For example, can it tell the difference between Web audio and Web video? Can't recognize BitTorrent traffic, regardless of the port it’s traveling across?
Quality of Service (QoS) complements packet shaping, and it often requires packet shaping to work effectively.
Packet shaping delays and limits low­-priority Internet traffic, which makes room for higher-­priority traffic. QoS, in turn, prioritizes the important or delay-­sensitive network traffic, making sure it gets sent out before other traffic streams and guaranteeing that it arrives at its destination within a specified time frame (e.g., under 100 milliseconds). Real-­time media, such as voice, video and online gaming, react badly to latency and jitter, so they often need QoS. For instance, latency above 145 milliseconds makes Voice over IP calls unlistenable. On the other hand, most of us can surf the Web at much higher latencies without noticing a serious delay, so there’s no point in applying QoS to ordinary Web traffic.
  • To obtain true QoS, with guaranteed latency and bit rates, you need to have some control over the entire network connection, including the equipment at both ends and all the circuits and routers in between. Usually, this means you have to pay your service provider for higher-­quality point­-to­-point connections (aka leased lines), because you can’t guarantee QoS over the public Internet.
  • On the flip side, you can implement packet shaping on one end of a network connection, so it’s often cheaper and less difficult to set up. Packet shapers can usually prioritize certain traffic streams, but they can’t guarantee delivery the way an end­-to­-end QoS connection can.
WAN optimization doesn’t discriminate between traffic streams the way packet shapers and QoS devices do. Instead, optimizers use a variety of techniques to strip out the inefficiencies and redundancies from network traffic. In other words, an optimizer speeds up all the traffic that passes through it.
Compression and caching are two basic techniques that optimizers use, but there are a variety of advanced algorithms that we won’t get into here. As with the other devices that we’ve discussed, WAN optimizers usually reside at the edge of your network, behind your router or firewall.
  • With certain types of WAN optimizers, you need a device at both ends of the network connection. For example, to take advantage of compression, you need a device at the far end of the connection that’s capable of decompressing.
Web caching is a piece of software or hardware that saves copies of recently accessed pages in memory or on a hard drive in order to speed up retrieval.
If you have 200 patrons surfing the Web on any given day, chances are that they’re accessing the same sites over and over again. Some of these sites are saved, or cached, in the Web browser on each individual desktop computer. If you return to a story you were reading half an hour ago, you don’t have to wait while your browser pulls down a fresh copy from a remote Web site. Instead, your browser shows you the version contained in the browser’s local cache. However, this only helps your personal connection. If the patron next to you wants access to the same article, they can’t get it from your local cache. They have to send another request to the remote Web server and download the same article that you downloaded half an hour ago. A Web­ caching server acts like a big shared cache for the entire network. All the PCs on the network can be configured to check the server first to see if a copy of the desired page has already been retrieved for someone else.
  • Some Web pages are constantly updated (e.g., weather information or stock quotes), and other sites are fairly static.
  • Obviously, your caching server shouldn’t hold and redistribute a page of stock quotes that someone retrieved two hours ago.
  • Any request for time­-sensitive information should bypass the caching server. Fortunately, most Web pages contain this information. A field in the page header will tell your caching server how long the page should be held.