The Return of the God Boxes

By Dave Roberts, Inkra Networks

July 19, 2002 -- (WEB HOST INDUSTRY REVIEW) -- Spring’s arrival always serves as a herald in the networking industry: it’s trade show time again. This year is no exception. While attending various networking trade shows and conferences this spring, I noticed an interesting trend: the “god box” is back. At least that was the scuttlebutt. Many vendors, old and new, seemed to have boxes of all shapes and sizes that integrated a wide array of technology, and all the various competitors took great pride in labeling each other’s products as god boxes. As somebody working on a networking infrastructure product that others have sometimes charged is a god box, I have found it interesting to see the industry wrestle with issues surrounding the integration of functionality.

Within the industry, there are three basic camps. The first camp, the god box camp, says that integration is good—the more the better. The second camp, the best-of-breed camp, says that devices should be as “pure” as possible, implementing a single functionality, but doing it very well. The third camp says that neither of the polar opposites is correct and that the right way to go is something in the middle, integrating some things and not others. Myself, I’m in the third camp. (In fact, most everybody considers himself or herself in the third camp. Like political parties, it’s cool to be in the middle.)

In this article, we’ll examine the first camp, the god box camp. We’ll try to understand what a god box is and what the strengths or benefits are of such an approach. A future article will examine the best-of-breed camp and try to understand the tradeoffs surrounding that approach. As we’ll find, there are merits to both approaches, and like most things, some disadvantages if things are taken to the extreme.

God Box Defined

So exactly what is a god box? As for most things that come in shades of gray, definitions are difficult. People have a hard time describing a god box, but “they know it when they see it.” Certainly, no equipment vendor would actually admit to building something called a god box. The term itself is pejorative, typically spoken of other companies’ products, not of one’s own. For the sake of argument, I’ll define a god box as any piece of networking equipment that attempts to do too much, and as a result ends up satisfying no one, and finally fails in the market place. This is in opposition to “best-of-breed” solutions that stick to the implementation of a single function and allow people to mix and match functionality, buying each function from a different vendor.

The theory is that it’s too difficult to be the best at everything. Therefore, best-of-breed solutions win in the market place, because they are able to build an unassailable technical lead. Not having a strong focus, companies working on god boxes aren’t able to compete.

That’s the theory, at least. Let’s look back in history and see if it holds up.

The first god boxes arrived when LAN switching was starting to become popular. They typically tried to implement support for multiple LAN protocol types, including Ethernet, token ring, ATM, and perhaps others. Often, these products were based on ATM backplanes, because ATM was designed to be the long-term convergence protocol. While previous-generation hub products had included support for multiple protocol types, these integrations typically used separate backplanes for each medium and did not attempt to interwork between them.

The attempt to harmonize all these protocols meant that the products were very complex and required large teams of people to build. Because of this, projects often missed schedule milestones and ran over budget. In and of itself, this is not new, as many high tech projects run late and turn out to be more expensive than originally thought, but these programs trended toward the worst end of the spectrum. As a result, many god boxes never saw the light of day. Often, those that did make it to market then failed to achieve success.

At least that was the perception. The reality is that these boxes failed before they were even built, because the product planning assumptions were wrong. The underlying theory was that everybody wanted to mix all these different protocols together all the time. These systems worked well if customers really needed to mix and match token ring, Ethernet, and ATM, but the fact was that most customers didn’t have that need. Most customers chose a fundamental technology and then stuck with it, making changes only rarely and under great market pressure (the conversion of token ring customers to Ethernet is still going on, years after it was obvious that Ethernet had won the LAN protocol wars). The fact was, you were either an Ethernet shop or a token ring shop, rarely both.

Even large corporations that did run both technologies typically limited deployments to a given workgroup, floor, building, or campus. With large islands of a particular protocol deployed, customers didn’t need a god box that was capable of mixing and matching features on a port-by-port basis. Customers only required a few internetworking devices (a router, for instance) capable of connecting these otherwise large, autonomous islands, located at strategic places in the network. The internetworking devices would be expensive, but there wouldn’t be that many of them. Cheaper, single-medium switches could be used to serve each media type.

By burdening god boxes with all the cost and complexity required to do everything, the god boxes couldn’t compete. This was a fatal flaw as the market chose Ethernet as the universal LAN technology and moved aggressively to drop prices and optimize around it.

So if the multi-protocol switch ended up being a god box, what is an example of a successful product that integrates a number of different features? The best example is the multi-protocol router. In contrast to multi-protocol switches, the multi-protocol router has been very successful. While customers did not want to mix and match protocols on a port-by-port basis, there was (and always will be) a need to mix and match protocols at some point in the network. The router is the ideal place to make this happen. Layer 3 switches have optimized the routing function for IP/Ethernet-centric environments and cut into the growth of the multi-protocol router segment, but there continues to be a need for a device that is smart enough to deal with multiple protocols simultaneously.

Finding The Value in Integration

So what’s the lesson for today’s market place? Are all these new devices really god boxes, or is that just mud-slinging from competing vendors? Like the two market examples we just examined, the answer is “probably a little of both.”

In the end, the key issue is whether there is any great advantage to putting various functions together in a single box, and whether that advantage could be delivered more optimally using some other strategy. If the features are thrown together randomly, just because somebody can build it that way, then the system is probably a god box.

On the other hand, if there is some tangible benefit to the features being packaged together (cost, convenience, management simplicity, and so on) then the product may have some merit. To know for sure, it’s important to look at customer usage patterns, and to see how many customers would benefit from the integration. If customers don’t want to use the various integrated features in combination, then the box is still a god box.

In a future article, we’ll examine the polar opposite to the god box: the best-of-breed solution.