EAGER: CA-RAM: Enabling Fast and Versatile Packet Processing for Future Large-Scale Networks

点击翻译按钮获取中文摘要

This exploratory research addresses the increasingly more critical performance, power, and functional requirements of search operations in key inter-networking devices like routers and firewalls. This project will investigate a specialized yet flexible search substrate called CA-RAM (Content Addressable Random Access Memory) that possesses great promise for high search performance at low power consumption. CA-RAM is based on the hashing theory and techniques. By decoupling dense memory array from compute components, CA-RAM achieves high storage density and configurability. This project will focus on how CA-RAM can effectively tackle all major packet processing applications, including packet forwarding, packet filtering, and deep packet inspection, in a scalable and complexity-effective manner. The project will also look at how new emerging memory and chip integration technologies can be used in CA-RAM to further demonstrate the effectiveness of the proposed approach. Intellectual Merit: This research has the potential for breakthroughs in the design of future inter-networking devices. The expected contributions include a directed exploration of the CA-RAM designs. Additionally, the researchers will gain an understanding of how the emerging 3-D chip integration and Phase-Change Memory technologies will further enhance the capabilities of CA-RAM and the packet processing devices built with CA-RAM.Broader Impact: This research will impact the design practices of inter-networking devices, a critical and essential infrastructure for our society. This project will help sustain the historical scaling rate of the network and Internet performance. The enhanced Internet will foster new exciting applications and services in global communications, commerce and entertainment, as well as cultural and technical advances. The researchers will document the approaches and findings in detail, and publish them in a timely manner.