HyperLens: Scalable High-Speed Packet Classification on Programmable Switches

Packet classification is a core operation in the network data plane. It is widely used in access control, traffic engineering, and firewalls. As network line rates and rule-set sizes grow rapidly, packet classification needs to support both line-rate forwarding and scalability. Programmable switches offer a promising platform. Their ability to perform high-speed parallel lookups in both TCAM and SRAM offers the potential to achieve scalability and line-rate forwarding. However, existing work struggles to achieve them simultaneously. Existing solutions fall into two categories. The first category encodes port numbers and compresses five-tuple rules in Ternary Content-Addressable Memory (TCAM), which limits scalability. The second category separates matching fields and offloads port matching to Static Random-Access Memory (SRAM). These schemes struggle to preserve semantic correctness and face their own SRAM scalability limits. In this paper, we propose HyperLens, which exploits the pipelining and parallelism features of programmable switches to achieve scalable high-speed packet classification. HyperLens leverages a divide-and-conquer idea to split five-tuple matching into three logical pipeline stages, presents a dependent-set prefix lookup algorithm and an independent-set range lookup algorithm to preserve semantic correctness across stages, and designs a TCAM and SRAM port-encoding scheme with parallel lookup to improve lookup efficiency. We implement and evaluate HyperLens on a P4 switch using large-scale real-world rule sets. Experimental results show that, compared with representative schemes, HyperLens reduces the average combined TCAM and SRAM memory consumption by 52%--62% across the evaluated rule-set scales, while sustaining line-rate packet forwarding.

Authors

Institutions

Publication Details

Journal
Proceedings of the ACM on Networking
Published
2026-11-19
DOI
https://doi.org/10.1145/3856967
Primary Topic
Network Packet Processing and Optimization
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

HyperLens: Scalable High-Speed Packet Classification on Programmable Switches

Shaoyong Guo, Xingguo Long, Kun Huang, Rongwei Yang et al.
Proceedings of the ACM on Networking
Network Packet Processing and Optimization
article

HyperLens: Scalable High-Speed Packet Classification on Programmable Switches

Shaoyong Guo, Xingguo Long, Kun Huang, Rongwei Yang, Zhiping Gao, Juzhong Wei
article en

Abstract

Packet classification is a core operation in the network data plane. It is widely used in access control, traffic engineering, and firewalls. As network line rates and rule-set sizes grow rapidly, packet classification needs to support both line-rate forwarding and scalability. Programmable switches offer a promising platform. Their ability to perform high-speed parallel lookups in both TCAM and SRAM offers the potential to achieve scalability and line-rate forwarding. However, existing work struggles to achieve them simultaneously. Existing solutions fall into two categories. The first category encodes port numbers and compresses five-tuple rules in Ternary Content-Addressable Memory (TCAM), which limits scalability. The second category separates matching fields and offloads port matching to Static Random-Access Memory (SRAM). These schemes struggle to preserve semantic correctness and face their own SRAM scalability limits. In this paper, we propose HyperLens, which exploits the pipelining and parallelism features of programmable switches to achieve scalable high-speed packet classification. HyperLens leverages a divide-and-conquer idea to split five-tuple matching into three logical pipeline stages, presents a dependent-set prefix lookup algorithm and an independent-set range lookup algorithm to preserve semantic correctness across stages, and designs a TCAM and SRAM port-encoding scheme with parallel lookup to improve lookup efficiency. We implement and evaluate HyperLens on a P4 switch using large-scale real-world rule sets. Experimental results show that, compared with representative schemes, HyperLens reduces the average combined TCAM and SRAM memory consumption by 52%--62% across the evaluated rule-set scales, while sustaining line-rate packet forwarding.

Proceedings of the ACM on NetworkingVol. 4(CoNEXT4)
Beijing University of Posts and Telecommunications (CN), Chinese Academy of Sciences (CN), Institute of Computing Technology (CN), Peng Cheng Laboratory (CN), University of Chinese Academy of Sciences (CN)
Network Packet Processing and Optimization
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon