PE WPP

Overview:

Posedge's Wireless Packet Processor (WPP) is a highly programmable Packet Processing engine that can be used to realize scalable, high performance, application aware Access Points/Base-Stations for Wireless LAN and LTE.

WPP routes multi-gigabits of data and provides sophisticated Quality of Service (QoS) with bandwidth shared across large number of clients offloading the Host Processor significantly.

WPP provides a Policy framework unified across wired and wireless Layer-2-Layer-4 across the LAN to support applications like Voice over WiFi and Video over WiFi, which require end-to-end QoS. In addition the Policy framework allows policies based on User-ID, Location etc.

The application aware engine supports identification packet flow for protocols like Skype, MSN and Bit-Torrent and applies configured QoS parameters to the identified flows.

The architecture provides very low response time latency. The latency of the Access Point to respond to a U-APSD trigger, defines the Power dissipation of the Client as the client initiates PS-POLL packet and waits for the data. The total time to respond is in the order of 100 us. Hardware based queues and data transmission right from queues, enables much faster latencies to schedule a packet to the client.

WiFi as it is specified is not reliable for multicast traffic (no ACKs for multicast). So opportunistically, the Access Points convert the multicast traffic to Unicast to make the communication reliable. WPP performs the multicast to unicast conversion with any buffer copies, by just maintaining reference counts in the hardware itself, there by avoiding the costly buffer copy operation by the Host Processor.

The WPP architecture guarantees wire-speed packet forwarding between the Gigabit Ethernet and the Wireless Port and is scalable with respect to the number of clients, number of flows, packet classification and packet editing requirements.

Deliverables

Host software required to setup, maintain and tear down the fast path tables. The modules adapted from the standard Linux kernel with net-filter hooks adaptation.

Net Interface driver

TMU Driver for adjusting QoS Parameters

Fast-path Code Classifier code (C-level) for the protocols implemented in the fast path engines. The PE code is currently available for the IPV4, VLAN, Bridge, ARP, RARP, IGMP, IGMP proxy protocols.

Features

Application Aware QoS.

Scalable to IEEE-802.11ac 3x3 data rates up to 1.5 Gbps line rate. Supports multiple flows (64K) and clients (256)

Fast path / slow path architecture. Scalable architecture based on performance requirements

Packet classification with multiple coherent 32-bit RISC processors. Supports Low Latency queues for Low Power Peripherals

Multiple SSIDs with different Security Levels. Complex functions like QOS and buffer management in Hardware

Per AC – Per STA Queuing and Shapers per queue. AMPDU-AMSDU aware scheduling of packets

Enables Air-Time Fairness across stations with different speeds. Proprietary internal bus for optimal throughput and highest performance with separate buses for data and packet communication.

IEEE 802.3 compliant Gigabit Ethernet Mac core with support for jumbo frames, VLAN tagging and flow control. Connects to Posedge WLAN MAC Controller and configurable to other MAC Controllers.

Flexible Packet store, with Internal and/or using Host DDR interface. Multicast to Unicast Support without buffer copy

Programmable Packet Editing. Optimal Hardware/software partitioning