Tuesday, March 9, 2010

DPI Deployments - Part3: Vodafone Hungary Case


It was probably a big surprise to Camiant when Vodafone Hungary (2.6M subscribers on Decmebr 31, 2009, #3 in Hungary) agreed to the press release they issued last year: Vodafone Hungary Deploys Camiant's Multimedia Policy Engine (here). Vodafone provided a quote to the release, and the release got nice coverage. A year after, Camiant still has a large homepage link to the release!

Camiant sells policy servers ("Multimedia Policy Engine"), to cable (PCMM architecture) and mobile (PCRF architecture) operators. PCRF is relatively new, and it seems that Vodafone Hungary is one of the pioneers in implementing the architecture.

In this case, the traffic control device is Starent ST-40 (now Cisco's ASR 5000) GGSN, who claims to have DPI (here).

The solution is describe as "Camiant's MPE, using 3GPP compliant interfaces in conjunction with Starent's ST-40 high capacity platform, provides Vodafone Hungary with dynamic bandwidth control for users that exceed their monthly limits. Traditionally, mobile operators have charged overage fees or stopped users from further using the network when they exceed a monthly usage cap. At Vodafone, Camiant's MPE enforces a moderated bandwidth level for these users but only during peak hours. This allows Vodafone's 3G users to get maximum value from their service but not at the expense of other users on the network."

Vodafone quote says "Mobile broadband is a critical part of our growth strategy," said Péter Török, Mobile Broadband Senior Manager of Vodafone Hungary. "Camiant's MPE technology allows us to differentiate our 3G services in the market, giving us a great competitive advantage and enables us to assure that our subscribers have the best possible end-user experience."

In the following diagram, Camiant shows how PCRF based policy server is implemented in a mobile network. Note that the policy server may control a number of network elements, including a GGSN (like our case here) and standalone DPI elements.




So while we don’t know exactly what policies Vodafone is implementing, and whether they control only subscribers or a combination of subscribers and applications, we see here a great example of an operator implementing a "fair-use" policy that limits few subscribers for the benefit of all others - and is brave enough to tell this to the world!

Monday, March 8, 2010

DPI Deployments - Part2: Why Hiding?


Operators are implementing DPI and Traffic Management solutions - it is a fact - as we know that DPI vendors are successfully selling and publishing those anonymous wins (see "part1: the anonymous riddle" - here).

So why operators proudly announce their infrastructure vendor selection (e.g. Verizon Wireless picks Ericsson, Alcatel-Lucent for LTE buildout or AT&T selects Alcatel-Lucent as LTE equipment supplier ) and are so shy when it comes to DPI?

Because of a relatively small number of very loud customers and the recent US Net-Neutrality initiative.

The following chart shows how, potentially, DPI may be used - in a varying degree of the problem it imposes to the operators:


In 10 years of existence, most DPI implementations were designed to handle congestion - operators could not supply the demand for bandwidth created by the extensive use of P2P filesharing. Once this was detected by the heavy file-sharers community - it became a "bad PR" to publish the deployment of traffic control - although it did well for the vast majority of subscribers (According to AT&T, the top 5% of their DSL customers use 46% of the total bandwidth).

So while using DPI for security, or regulation is either good or a must, and using DPI for profiling subscriber behavior is a big No-No (companies is this space disappeared – see: NebuAd grilled over hot coals in Congress on privacy) - the most common use of DPI should be good, and is even supported by the current Net Neutrality guidelines (under "reasonable network management") - but operators are still afraid of telling their customers they do it.

But we do have some brave operators.

To be continued.