1-800-HTTP
#mobile Trend toward applications as value added service requires application fluency. A newish trend is quietly (comparatively speaking) increasing across the Internets. The basic premise is that end-users aren't being charged for data traversing mobile networks when accessing certain sites, much in the manner users are not charged for calling a company's toll free (800) number. For some companies, this largess is intended only when accessing their site. More recently, at least one provider (T-Mobile) has gone to making free access to other sites, like Facebook. AT&T to Let Content Companies Subsidize Users' Data Costs AT&T's new service, called Sponsored Data, works like a 1-800 number but for data. In theory, a company like Netflix Inc. could promote a new series by covering the cost of data that otherwise would count against a subscriber's plan. Subscribers will see a logo that lets them know the content they are accessing doesn't count against their data plan, and sponsored data volumes will be broken out on their bills. This month, T-Mobile US Inc. began giving subscribers to its lower-cost GoSmart Mobile prepaid service free wireless access to Facebook, even if those customers aren't paying for a mobile-data plan. Unlike with AT&T's new plan, T-Mobile is bearing the data costs under the offer, which could bring new customers to the service and get lower-end subscribers into the habit of using mobile data. Now, aside from this being a Very Cool Idea from a consumer perspective, consider what such offerings require under the hood to actually get implemented. The provider has to know the destination and whether or not that particular destination falls under what I'm going to call a "1-800-HTTP" plan. In order to know the destination, the provider must be able to look inside the data, the payload, of mobile network traffic. It must be able to, in real-time, inspect that traffic, extract the HTTP headers, and evaluate the host name against its list of "1-800-HTTP" plans. In real time. Certainly the provider could use IP address destinations instead. Let's face it, Facebook's block of addresses probably doesn't change all that often. The same goes for most established organizations with a presence on the Internet. The problem with this approach is the same as its always been when we start talking about whitelisting (or blacklisting) by IP address. It's not necessarily that they'll change that's problematic, it's the sheer number of IP addresses that must be maintained in a table and, ultimately, searched in real-time. Geeks will recognize this as the "IPtable" performance problem inherent in the proliferate use of iptables for software implementation of a variety of network services, often firewalling and security-related services. The longer the list of entries in the table, the greater latency is introduced when switching and routing requires evaluation of a destination (or source) IP address against that table. Thus, while early implementations of the "1-800-HTTP" plans may very well simply use IP addresses, if such offerings generate enough interest and grow rapidly, an IP address-based system generally will not scale. IT'S ABOUT MORE THAN A HOSTNAME But that's not the only thing going on under the covers. If you read the announcement, you'll note the statement "Subscribers will see a logo that lets them know the content they are accessing doesn't count against their data plan." In other words, an advertisement. Content. Application content. That content must be inserted into the payload, into the HTTP (aka application) payload. That means deep content inspection and modification, which requires visibility into layer 7. That means some kind of service has to exist in the data path that is capable of interacting with application data, that is fluent in the language of the application. That fluency is required in order to insert into its data this "logo" or advertisement expounding upon the largess of the organization sponsoring the free access. What this all means is that as service providers continue to seek out new ways to add value to their (admittedly already valuable) networks, they will necessarily continue to move "up the stack" toward the application. There are a plethora of value added services based on application and data that could be made available by service providers, provided they have the right infrastructure in place. That infrastructure must be not only high-performance, but it must be able to maintain that performance while executing layer 7 (application) fluent-based services in the data path. Whether services like "1-800-HTTP" will continue to gain traction and lead to other application / data specific offerings depends on whether or not mobile providers can maintain performance while executing some fairly intensive inspection and modification under the covers.