I have been around the world (in some cases, in a single trip!) to meet with existing Procera customers and interested prospects on our solution. During a recent trip, a coworker made an offhand comment after we met a customer that stuck with me. “The Internet is a big place.”

The moment he said that, it really brought home something to me that had been creeping into my thinking during the trip. During the trip, I met with customers, prospects, resellers, and analysts in several different countries, and there was one consistent theme that I took away from those meetings: Every customer is a little bit different. Different parts of the world have different needs, and although the network diagrams may be similar, just as in a city, different neighborhoods have different needs.

In the context of what I do for Procera (Product Management), what that means is that every network, every customer, every service offering, is a little bit different. Normally, in Product Management, your hope is that every customer needs exactly the same thing (which makes your life much easier when defining a product). Unfortunately, our Intelligent Policy Enforcement technology has so many different uses that we see variations of even some of the basic use cases. It also means that product specifications and network demands may mean that a product that appears to line up in basic specification is actually not the right solution for a customer.

For example, let’s look at an often ignored (or minimized) specification on application aware systems – session capacity and setup rate. All around the world, we are seeing a sharp increase in session usage as the era of “always-on” social networking takes effect. Mobile network session usage, which used to be minimal, has skyrocketed with the rise of iPhones and Android phones that are multitasking and constantly connected to the network. Applications and sites like foursquare, Facebook, Mixi (Japan), Renren (China), Twitter, and IM/VOIP clients are increasing the consumer’s urges to stay connected at all times to the Internet. In Asia, networks have very high session setup rates per user, mainly due to the proliferation of file sharing clients and their behavior. In Canada, the session usage per user is almost three times the normal usage in the United States due to more file sharing activity. I have seen a number of deployments fail due to session loads that overwhelm stateful aware systems, and not just during DOS attacks, but in basic system operation.

Another key metric is the number of subscribers supported on a single system. Across fixed and mobile networks in Europe and in fixed networks in the America, the number of subscribers per system that needs to be supported is lower due to the number of operators competing in a single geography. In mobile networks in Asia and the Americas, that number can skyrocket (think of the mobile subscriber counts for Verizon, ATT, NTT, and China Mobile) to millions per system. Understanding how systems can grow or scale subscribers and policy management deployments are critical, since as the subscriber count grows, so too does the signaling load from Radius and PCRFs. A system that supports 10M subscribers, but only 1000 transactions per second will fail in a deployment since it cannot maintain the provisioning rates required for subscriber management.

And then there is a key metric for deployments: reporting and visibility. As the number of subscribers and the amount of sessions increase, the amount of statistics collection and storage required by the system shoots up as well. The amount of traffic does not matter as much in this scenario as you might think, since a single subscriber with one application is really just one statistic item, but one subscriber with 1000 applications, even at very low rates, is 1000 statistics operations that need to be collected, written to disk, and then analyzed. In North America, network operators are very interested in the impact that Netflix and other over the top video has on their network, and they want that information collected not just as “video”, but are interested in what devices are used, what providers are popular, trend information over time, and many other variables so that they can better plan their peering strategies. Mobile operators want to understand the trend and applications for the use of social networks, as they want to offer service plans that are attractive to avid Facebook users. In South America and Asia, file sharing and YouTube are huge consumers of bandwidth, and operators want to make sure that their capital expenses for peering are not spent entirely to support this traffic types, and want to send that type of traffic over lower cost transit links.

Not to mention the biggest product question of all – what features are going to be activated? In many products, the more you do, the less performance and capacity the product has. This is especially true in the policy enforcement arena, and some products literally slow to a crawl when you activate certain functions. Although this is an expected result sometimes, it is important that when you are publishing your product specifications and testing products, make sure that you test it in the same way (and if possible at the same scale) that you intend to deploy it, as you may be surprised in some cases what happens when you move from the lab into production – sometimes unpleasantly so. It is very normal to have a system running in tap mode just as if it was going to be run on the production network to determine the real capabilities and performance of the system under load.

So where I am going with all this?

The Internet is a big place, and even though it is one Internet, it has neighborhoods that behave very differently, and make sure you buy the right product for your application. Every network is different, even if they look like the same network diagram. I certainly think that in most cases, Procera’s product can get the job done better than others, but at the end of the day, a network operator has a business to run, and their job is to make that network run as efficiently as possible, for their configuration. So before you buy, make sure you are getting what you want, don’t just buy because of numbers on spec sheets. Customer’s should behave as the old Burger King slogan used to go “Have it your way” – because your needs ARE different than other customers.