4/30/2002 -- You know a topic is getting hot when you're at a party and a friend who is into Novell networking with a bit of Microsoft says, "So, tell me about Content networking."

Content networking doesn't refer to a single product or product line or even to a particular task. It refers to being able to get content to the user as fast as possible with a minimal amount of fuss. There are some content devices that are separate boxes and others that can be routers running extra services. It all depends on what you want to do.

While I've been unemployed, I've taught several classes in Asia. When I access the Cisco Web site from home, traffic has to bounce around from provider to provider before hitting www.cisco.com in San Jose. When I taught in Singapore or Hong Kong though, it didn't make sense to have all of my requests cross the Pacific. A content networking device called a Distributed Director can reroute my request to a Web server much closer to where I am, say Tokyo, reducing latency and bandwidth usage.

There are a couple of different ways to do this. First, through the use of recursive requests, the initial DNS request for a given domain will be forwarded to a device that figures out where the connection should be rerouted to based on Director Response Protocol (DRP) metrics. The second method is for the content device to be the "www.somewhere.com" device. It knows what servers can serve pages for www.somewhere.com and based on the DRP metrics configured, it does an HTTP redirect.

So, what is DRP? DRP is a protocol that figures out how far a given location is based on routing information. In external mode, it uses BGP autonomous system numbers to figure out distance on a hop-by-hop basis. One AS equals one hop. Server mode uses the metric used by an IGP to determine which is closest. Random mode just randomly selects servers in a pseudo-load balancing fashion. Other modes exist, but those are the major ones.

One downside to networking on the Internet is that depending on how you look at it, routing may or may not make sense. For example, when I try to access www.cisco.com, VisualRoute shows my path going from Phoenix, to Atlanta, to New Jersey, to either Los Angeles or Dallas (depending), back to New Jersey then over to San Jose. When configuring the Distributed Director, you can account for paths like this by using the DRP Client-to-Server Round-Trip Time Metric. The server with the least amount of latency would handle the request.

In the event that outside distribution and load balancing isn't needed, there are lower cost devices that can do roughly the same thing without taking BGP into account. The Local Director is used to reroute traffic for load balancing and optimization purposes but is used inside a single autonomous system. This means you can use a Local Director to reroute requests for intranet access to the server closest to the user, rather than using slow WAN links for all requests to go to the headquarters.

The 11000 series Content Switch is a different beast, designed to speed up secure connections. SSL is a convenient method of serving secure Web pages but it's rather processor intensive, and many of the servers used to host Web pages aren't optimized to handle secure traffic quickly. The 11000 series is. SSL processing is offloaded to the content switch, allowing the Web server to serve Web pages. This is a great box for e-commerce and online transactions.

Content devices vary in what they can do. Some of the devices speed up content; others, like IP/TV devices, actually serve content. If you're involved in supporting a high-traffic Web site using SSL or are considering multiple hosting providers, I recommend you do more research into these devices. You'll make your customers much happier.