I saw today that
Cablevision du Nord (Québec, Canada) is using DPI (
here) to "To deliver the best possible viewing experience for our subscribers".
The story says that "
Cablevision joins nine of the top 10 U.S. operators and three of the top five Canadian operators to deploy the solution." so I thought that my 5th piece on DPI deployments should cover North-American cable operators - which brought DPI to the public attention.
I have seen the DPI industry developing over the last 10 years in many aspects. Starting with how we call this space - in the earlier days the more common name for today's DPI/Traffic control solution was "
Bandwidth Management", "
QoS" or "
Policy based Networking". Actually we should consider going back to the latter - as it fits nicely the current architecture trends (using PCRF and PCEF).
Another interesting aspect is the deployment trends – identify the early adapters and followers
Around year 2000, when broadband (over DSL, cable, fixed-wireless and satellite) became affordable and widely deployed, and the music download (Napster et al) showed up, service providers started to notice the congestion it creates and costs required to maintain adequate quality of experience to their subscribers. Their access network and uplinks became congested, and just adding more bandwidth did not help, as it was all used shortly after installing. See more on my post - "
The DPI Story – Part II – DPI for Traffic/Bandwidth Management" (
here).
Naturally, the earlier adapters were a combination of:
- ISPs that see significant downloading traffic (US was the first)
- Smaller ISPs with limited uplinks capacity, and tight budget
- ISPs operating in regions where bandwidth is expensive (Asia, Latin-America)
The bigger providers - incumbent DSL providers for example - were (and still are) reluctant to deploy traffic management solutions. There are exceptions, of course, but even before Net Neutrality became an issue, many of them saw themselves as carriers - carrying bits that are all made equal.
In a very general view the general global trend over time was - small ISPs (IOCs, fixed-wireless, local cable operators) in North America -> ISPs of all sizes in Latin America and Asia -> ISPs in Europe and in the recent 2 years mobile operators. One major exception is the North-American MSOs (cable) that started to implement DPI relatively early compared to other service providers of their size.
The reason for the MSOs to use DPI may have been the fact that their "last mile" to the subscriber premises is shared by a large number of subscribers (similar case exists in wireless networks) which may cause local congestion situations, which may lead to dissatisfied subscribers and churn. The alternative- splitting CMTS - may have been too expensive, compared to a central DPI/QoS solution.
Sandvine, focusing on the North-American MSO market, managed to capture most of it. It is assumed that Comcast alone generated, over several years, over $60M for Sandvine. Unfortunately, Sandvine’s technology for traffic shaping (i.e. limiting bandwdith for P2P file sharing applications) involves the use of TCP RESET commands, which are visible to the subscriber device.
Once this was detected (
here), it led to the FCC action against Comcast. While the effect of a RESET packet may be described as blocking traffic, it is actually a way to limit the subscriber download speed to a certain level, and not a complete block. Other shaping techniques, like queuing, provide the same effect but is invisible to the end-user.
So today almost all MSOs in North-America use DPI solutions – we just need to wait and see what limitations the FCC will impose. So far it seems that two of the FCC already stated guidelines – disclosure of service terms and “reasonable” network management to reduce congestion and maintain quality of service” could resolve the conflict.