Elastic Switch Programming with P4All

Abstract

The P4 language enables a range of new network applications. However, it is still far from easy to implement and optimize P4 programs for PISA hardware. Programmers must engage in a tedious "trial and error" process wherein they write their program (guessing it will fit within the hardware) and then check by compiling it. If it fails, they repeat the process. In this paper, we argue that programmers should define elastic data structures that stretch automatically to make use of available switch resources. We present P4All, an extension of P4 that supports elastic switch programming. Elastic data structures also make P4All modules reusable across different applications and hardware targets, where resource needs and constraints may vary. Our design is oriented around use of symbolic primitives (integers that may take on a range of possible values at compile time), arrays, and loops. We show how to use these primitive mechanisms to build a range of reusable libraries such as hash tables, Bloom filters, sketches, and key-value stores. We also explain the important role that elasticity plays in modular programming, and we allow programmers to declare utility functions that control the relative share of data-plane resources apportioned to each module.

Safe, modular packet pipeline programming

“…While Lucid2 was motivated by the constraints of PISA architectures in general, and the Intel Tofino in particular, pipelined parallelism is a widely-used technique for improving the throughput of data-processing applications. Lucid2's design and type system may provide a guide for future researchers looking to deploy these ideas in the context of other network devices (Baldi [2020]; Kalkunte [2019]), other network programming languages (Gao et al [2020b,a]; Hogan et al [2020]; ), or other domains entirely, such as signal processing (Ebeling et al [1996])…”

“…More recently, there have been a number of efforts to make programming P4 switches easier. For example, Domino ), Chipmunk (Gao et al [2020a]), Lyra (Gao et al [2020b]), and P4All (Hogan et al [2020]) allow programmers to use high-level, imperative, C-like languages to describe switch computations. They then deploy program synthesis techniques to allocate those computations to stages in the pipelines of one or more hardware devices.…”

“…We view Lucid2's contributions to this space as complementary to, and synergistic with, these other effortsÐone can certainly imagine future systems in which programmers are constrained by Lucid2's type system, ensuring computations can be compiled, and use synthesis techniques to spread computation across one or more devices. Indeed, Lucid2's vectors and loops were inspired by related unsafe features in P4All (Hogan et al [2020]). By incorporating appropriate elements of Lucid2's type system, P4All could deliver safe vectors, loops, and synthesis in the future.…”

An Exhaustive Survey on P4 Programmable Data Plane Switches: Taxonomy, Applications, Challenges, and Future Trends

“…A noteworthy work (P4All [353]) proposed an extension to P4 where operators write elastic programs. Elastic programs are compact programs that stretch to make use of the hardware resources.…”

Lucid: A Language for Control in the Data Plane

“…The optimizations in Lucid's compiler are heuristics based on our experiences with hand-coding efficient P4-Tofino programs. Recent work suggests that sophisticated optimizations based on synthesis [10] or ILP [14] algorithms may do better. However, in general, finding an optimal PISA layout is NP complete [32].…”