Sorry, you need to enable JavaScript to visit this website.

Faculty Awarded Grant from Cisco

Faculty Awarded Grant from Cisco

Dr. Minseok Kwon, Associate Professor of Computer Science, along with RIT colleagues Bruce Hartpence from the Information Sciences and Technology Department and Gahyun Park from the Department of Computer Security, recently received a corporate research grant from Cisco.

The title of the project is: A length-aware cuckoo filter for packet forwarding at routers with the abstract given below.

The support is for one year with possible extensions and will involve building and testing a with Cisco UCS servers with Intel Xeon processors over 100Gbps high-speed network connections.

Abstract:
A cuckoo filter is a compact data structure for membership test allowing item deletion without filter reconstruction and extra space overhead. We will introduce a new filter called a length-aware cuckoo filter (LACF) that treats incoming IP addresses differently based on the popularity of their prefix length. Using different numbers of hash functions based on length popularity is formerly explored by length-aware Bloom filters. IP addresses of less popular prefix lengths are inserted into the filter twice, instead of once, to lower false positive rates. This reduces the false positive rates for the address of less popular prefix lengths, but does not increase the rates for the other addresses. Our analysis and preliminary numerical results show that the overall false positive rates decrease compared to the case when the original cuckoo filter is used. As proof-of-concept, we will implement LACF using Cisco’s VPP (Vector Packet Processing) library that runs on Intel’s DPDK (Data Plane Development Kit) adopted by FD.io. Our focus will be on the ip6-input node in VPP, and then extend it to the L2 input ACL. The system will be tested on both functions and performance. In the functional testing, we will verify the accuracy of forwarding results, and also measure the false positive rates. In the performance testing, we will measure forwarding table lookup time and throughput as the amount of data processed per unit time, and evaluate LACF against the latest FD.io VPP code release.