Internet Machines' NPE10 is the first single chip, fully software programmable, massively parallel network processor delivering full duplex 10 Gbps and OC-192c
speeds and scalable to OC-768c. The NPE10's packet processing capability is unmatched by any other network processor in the market today, enabling applications such as high-speed switching and routing, flow-based traffic policing, QoS-aware traffic classification and prioritization, and packet header modifications to be implemented more rapidly than with other solutions.

Simply Powerful
The NPE10 offers full software programmability with no performance penalty. Its innovative, massively parallel architecture is based on embedded RISC processor cores, and is driven by an easy-to-use, single-threaded, single processor programming model.

Simple. The NPE10's unique architecture and hardware-based single-threaded programming model allow system developers to use familiar and conventional programming techniques to rapidly develop and optimize their application-specific packet processing forwarding code using either C or assembly languages. Developers have the flexibility of using Internet Machines' powerful, user-friendly Development Workbench software development kit, as well as industry-standard GNU tools.

Powerful. As a highly integrated, single-chip solution, the NPE10 executes sophisticated packet processing and forwarding functions in both ingress and egress directions, for an aggregate throughput in excess of 50 million packets per second.

Highly integrated. A single chip delivers comprehensive packet processing capabilities on both ingress and egress directions at 10 Gbps and OC-192c data rates.

Concurrent protocol-independent, multiservice support. "Any header and payload" processing enables the NPE10 to concurrently handle multiple different protocols and services on different subchannels, which increases system flexibility and breadth of applications.

Wire speed. The NPE10 maintains full duplex 10 Gbps wire-speed performance for all packet sizes.

Software programmable. As network service requirements change, the packet forwarding software code running on the network processor can be quickly upgraded in the field, eliminating costly hardware downtime or forklift upgrades.

Comprehensive software development kit. Internet Machines' Development Workbench includes a clock cycle-accurate software simulator of the NPE10, along with a library of sample application code and full suite of tools for code development, debugging, optimization, traffic generation, and performance analysis.

Features and Benefits Allows processing of multiple or different protocols within one aggregate data stream, including IP, MPLS, Gigabit Ethernet, ATM, SONET SPE, iSCSI, and more Maintains wire-speed performance for packets of all sizes at 10 Gbps throughput for both channelized and clear-channel OC-192 applications Provides a high-performance classification engine, using off-the-shelf ternary CAMs or other co-processors Supports OIF SPI-4 Phase 2 Interface Implementation Agreement, allowing for glueless connectivity to 10 Gbps/OC-192-capable framers, MACs, or traffic management devices Optionally connects directly to Internet Machines' TMC10 traffic management co-processor to enable a full-function 10 Gbps/OC-192c line card that is area and performance optimized Incorporates a high performance 64-bit/66 MHz PCI interface for configuration, management, and other control plane operations by an external host processor In applications that require additional perfor-mance, two NPE10s can be connected together in series or in parallell. Features NPE10 Other Solutions Highest integration: 10 Gbps single chip, full duplex yes half-duplex Greater than 50 mpps aggregate throughput yes no Fully programmable software processing at line rate yes no Protocol-independent: handles different protocols on different subchannels concurrently yes partial Simple to program yes no NO multi-threading, context switching, parallelism issues to deal with yes no NO proprietary programming languages or development tools to learn yes no Field upgradable with no loss of service yes no Channelization down to STS-1 rate ports yes no Cascadability for applications requiring additional horsepower yes no Seamless connectivity to protocol-independent traffic manager and switchfabric solution yes no