When Napster first appeared nobody knew the effect it would have on network traffic. Popularized by its music sharing ability, Napster epitomized a new breed of file sharing systems known as peer‐to‐peer or more commonly referred to as P2P. While the exact fraction of Internet traffic consumed by P2P has been much debated (one oft quoted claim was greater than 60%), it is undeniably large. According to a forecasting study conducted by Cisco, although P2P is declining as a percentage of overall IP traffic, it continues to grow in volume. The study estimates that “P2P file‐sharing networks are now carrying 3.3 exabytes per month and will continue to grow at a moderate pace with a CAGR of 18 percent from 2008 to
2013.”
Whatever the actual percentage, the challenge is that this P2P traffic is disproportionate and not revenue generating. Efforts to limit bandwidth based on TCP port numbers are now often sidestepped so a more discriminating strategy is required.Utilizing DPI (Deep Packet Inspection) techniques it is now possible to build an application platform that can identify and isolate specific P2P traffic, e.g., BitTorrent. Once identified, rules can be applied to control traffic based on a variety of variables. For example:
• Traffic type
• Specific User (IP)
• Specific Application (by user) e.g. P2P vs. HTML, VoIP etc.
• Traffic flow (direction)
• Tariff or rate class by user and/or traffic type In this way the P2P traffic may be throttled or diverted and priority given to more “sensitive” applications. Traffic management platforms have existed for some time although they were near totally proprietary and based on custom configured hardware. There were no “open” interfaces and one was tied to a single major
manufacturer. The reliance on a single supplier kept costs high, both for the development of new traffic software and the systems themselves.
The flexibility of an open customizable platform enables application developers to create software for a broader market and customer base. An environment where new traffic modules (both hardware and software) can be added easily, facilitates reuse and keeps costs under control. Advantech offers such a system platform that delivers maximum benefit for the software developer as well as the network OEM.
The NCP‐5260, based on a dual Intel® Xeon® 5600 Series motherboard, is connected to two Netlogic XLR Network Processor Boards performing deep packet processing on multiple 10GbE ports prior to forwarding packets to the Intel® Xeon® processors. This hybrid design has numerous performance advantages where standard x86 software can be implemented on the motherboard and packet processing offloaded to the NPU’s.Advantech’s NCP‐5260 is a clear choice for such demanding traffic management applications.
NCP-5260 for P2P Traffic Management
The NCP‐5260 represents a new generation of hybrid system designs with Intel® architecture processing on the control plane, and Packetarium™ network processing boards for the data plane. It integrates up to two powerful, multi‐core Packetarium™ network processing boards for wire‐speed packet processing and accommodates up to 16 x 10 GbE external interfaces. The main carrier board provides the high‐speed switched interconnects between Packetarium boards. The Intel Xeon‐based server board provides storage, system‐management and remote management network connections.
Each network processing board is linked by dual or quad XAUI ports to a Broadcom 10 GbE switch on the carrier board. The 10 GbE switch provides sixteen front panel 10 GbE SFP+ ports. The carrier board incorporates a MPC8545 processorfor overall switch management. A SATA controller on the server board connects to two 2.5" SATA HDD slots.
The scalability of the NCP‐5260 positions it ideally for OEMs designing high bandwidth systems in enterprise networking. It is particularly applicable for applications in service‐provider networks for enhanced security, in content‐aware routing and subscriber‐based services.
The initial Packetarium™ network processing boards supported by the NCP‐5260 are based on the Netlogic RMI XLR 732 8‐core processor. Each processor supports up to 4 GB of memory on two DIMM sockets. Two PCIe x4 provide control plane connectivity with the carrier while two XAUI ports connect to the data plane. The board is designed with IPMI 2.0 H/W management, remotely managed via a local Module Management Controller (MMC) connected to the carrier's IPMB‐L (I2C) bus. A console port and a 1000 Mbps port provide further management interface options. Other network processing boards in thePacketarium™ family are also compatible with NCP‐5260.
• Packetarium™ Network Processing Board design
• Hybrid design for Intel architecture on control plane, and Packetarium Board as data plane
• 6WINDGate support provides accelerated Fastpath packet processing offload without changes to legacy x86 code. Up to 7x more performance over the standard Linux network stack.
• 1 to 2 Packetarium™ Board slots
• Dual Intel® Xeon® 5500/5600 series support
• 16 x 10 GbE external interfaces
• 10 GbE switch w/ L2 switch management
• One standard PCIe expansion slot
• IPMI 2.0 HW Management
• Linux support
• FIPS Level 2 compliant
“Advantech’s NCP‐5260 Packetarium™ design is a clear choice for demanding traffic management applications requiring accelerated packet processing and Intel® server class data processing.”