Virtual Queues for P4: A Poor Man’s Programmable Traffic Manager

Abstract

The advent of programmable network switch ASICs and recent developments on other programmable data planes (NPUs, FPGAs) drive the renewed interest in network data plane programmability. The P4 language has emerged as a strong candidate to describe a protocol independent datapath pipeline. With its supported architectures, the P4 language provides an excellent way to define the packet processing and forwarding behavior, while leaving other networking components such as the traffic management engine, to non-programmable fixed function elements, based on the capabilities of most programmable devices. However, network flexibility is essential to meet the Quality of Service (QoS) requirements of traffic flows. Thus, enabling programmable control for fixed-function elements like traffic management is crucial.Towards that end we propose the use of virtual queues in the P4 pipeline, investigate the application of virtual queuebased traffic management, and portability of the approach using different P4 programmable targets. Specifically, we focus on virtual queue based Active Queue Management (AQM) for congestion policing and meeting the latency targets of distinct network slices. The solution is compared to P4 built-in functionality for bandwidth management using meters, proving also that the additional dimensions of control are achieved without compromising the processing complexity of the solution.

The potential of SDNs and multihoming in VANETs: A Comprehensive Survey

“…The proposed method uses virtual queues in the P4 pipeline for AQM, demonstrating how more control dimensions can be achieved without compromising processing complexity [157].…”

Qvisor

“…Virtualizing programmable networks Previous works have focused on virtualizing programmable networks, from their end-host counterparts to the in-network resources [6,7,15,17,18,24,34,[36][37][38][41][42][43]. While most solutions target software switches, SmartNICs or NetFPGAs [17,24,34,41], a few focus on hardware switches [16,38,[41][42][43].…”

“…Virtualizing programmable networks Previous works have focused on virtualizing programmable networks, from their end-host counterparts to the in-network resources [6,7,15,17,18,24,34,[36][37][38][41][42][43]. While most solutions target software switches, SmartNICs or NetFPGAs [17,24,34,41], a few focus on hardware switches [16,38,[41][42][43]. Among them, [16,41,42] optimize resource sharing at compile time by combining multiple applications into a single program, [38] enables runtime reprogramming of switch data planes, and [43] facilitates the dynamic sharing of switch resources across applications.…”

“…Among them, [16,41,42] optimize resource sharing at compile time by combining multiple applications into a single program, [38] enables runtime reprogramming of switch data planes, and [43] facilitates the dynamic sharing of switch resources across applications. Only [7,17] address the virtualization of the scheduling infrastructure of programmable switches, but they do not virtualize programmable scheduling policies.…”

Heavy Hitter Detection on Multi-Pipeline Switches

“…Even information about single flows may be spread onto multiple pipes. This may be due to failures, load balancing, and traffic engineering [5,21]. A given flow may start in one pipe and during its life-cycle it may move to others.…”