Monday, June 11, 2012

[Guest post]: How to Leverage the Multicore Technology for Setting up the Network Monetization

By David Le Goff*, 6WIND

It has been a thought for a long time but now it became true. The multicore platforms are strong enough to be used for network monetization. If you want to cook, here is the recipe:

Trust in the technology that is now being adopted in the network

Although the multicore technology has been around for many years, from an operational perspective justifying the business benefits we were still in an early stage until advanced packet processing methods leveraged the technology.

Today, all processors used in high-performance networking products are based on multicore architectures. These platforms provide the ideal environment for implementing the high-performance packet processing that is required for network appliances. For developers of such equipment, however, selecting a multicore processor for their system is only one step in designing a high- performance system solution. Generally, the more complex question is how to architect the software which typically needs to process packets from multiple streams of network traffic at wire-speed. A standard networking stack uses services provided by the Operating System and is subject to significant overheads associated with functions such as preemptions, threads, timers and locking. These processing overheads are imposed on each packet passing through the system, resulting in a major performance penalty for overall throughput.

A superior solution is provided by specialized packet processing software optimized for multicore architectures. In a well-designed implementation, the networking stack is split into two layers. The lower layer, typically called the fast path, processes the majority of incoming packets outside the OS environment and without incurring any of the OS overheads that degrade overall performance. Only those rare packets that require complex processing are forwarded to the OS networking stack, which performs the necessary management, signaling and control functions.

Get a high network performance environment

Tap network designs for monitoring, charging or writing statistics are good enough but once you tend to monetize the network and dynamically shape the traffic, inline processing is mandatory and asks you to get the highest network performance numbers for a business justification. With the latest improvements from the silicon leaders bringing on the market frameworks/SDK to make the most of the multicore platforms, software network companies have been able to achieve performance expected for the network inline processing operations. Available either for userland or bare metal implementation, these silicon starter kits have been used to design and develop complex network protocol stacks to get the following result examples needed for today’s network topology: 14M Packet Per Second (PPS) of forwarding on 2 cores, 5Gbps of IPSEC per core, 2M TCP concurrent sessions per CPU, 3MPPS of Layer7 DPI analysis per core enabling you to set up a network monetization environment at a justified cost.

Address the high flexibility requirement

With the constant evolving network landscape associated with the unavoidable regular marketing promotions advertised by the service providers for being either a premium, a freemium or whatever consumer, the networking equipment have to continually upgrade and this is all the benefit of using multicore platforms allowing pure software deployments, Linux-based for instance to match these expectations. Low power consumption, high memory bandwidth and crypto security processor supports are all additional ingredients to address the overall challenges.

Use the Massively Parallel Processing (MPP) capabilities

With the increasing data available in the networks, the analytic capability along with the execution capability were not able to follow the data growth and as a consequence, monetizing the network for an higher Quality of Experience (QoE), enabling the service providers to finely differentiate their end- users/subscribers profiles based on their network consumption/behavior was simply not possible.

This has been reached thanks to the multicore technology associated with advanced packet processing methods for processing inline the traffic while sustaining the high performance expectations. Combining a DPI-based engine with a smart flow table and traffic shaping mechanisms hosted on a multicore platform, answer to these QoE expectations.

Highly parallelizable and optimized for traffic management, this is the kind of solutions adopted by the networking equipment manufacturers to implement network monetization engines. The “shared nothing” Implementation thanks to the multicore technology associated with advanced packet processing methods will allow you to avoid memory bus congestion, locks, preemptions, etc…

Get it and Shake it

As you got all the secret sauce to know how to parallelize the network processing over multicore platforms, you are now able to build up a network monetization engine capable to proceed Layer 7 DPI-based analysis along with its associated traffic shaping and steering facing millions of consumer connections whereby the service providers will continually promote personalized subscriptions.
*David Le Goff is 6WIND’s product marketing manager, responsible for cloud and virtualization initiatives, working on product roadmaps, market trends and pricing strategies. He also manages 6WIND’s marketing activities in Europe.

David has 10 years’ experience in the telecom industry. His background includes leadership roles in both engineering and marketing at Alcatel-Lucent where he was in charge of the business development and the world wide Telco support of the Cloud services initiatives. Prior to these activities, David managed the carrier application data base business around IMS and LTE along with the first steps in the Business Intelligence domain for Telco, leveraging the data monetization. He began his carrier with Web services and Voice over IP R&D activities both at Orange Labs and Siemens.

No comments:

Post a Comment