numpywren: serverless linear algebra

TL;DR

numpywren demonstrates that serverless architectures can efficiently perform large-scale linear algebra operations with elastic scalability and ease of management, achieving performance close to traditional supercomputing solutions for certain algorithms.

Contribution

The paper introduces numpywren, a serverless system for linear algebra, and LAmbdaPACK, a language for parallel algorithms, enabling scalable and fault-tolerant computations.

Findings

Performance within 33% of ScaLAPACK for key algorithms

Up to 240% better CPU-hour efficiency due to elasticity

Limitations in network efficiency affect some algorithms

Abstract

Linear algebra operations are widely used in scientific computing and machine learning applications. However, it is challenging for scientists and data analysts to run linear algebra at scales beyond a single machine. Traditional approaches either require access to supercomputing clusters, or impose configuration and cluster management challenges. In this paper we show how the disaggregation of storage and compute resources in so-called "serverless" environments, combined with compute-intensive workload characteristics, can be exploited to achieve elastic scalability and ease of management. We present numpywren, a system for linear algebra built on a serverless architecture. We also introduce LAmbdaPACK, a domain-specific language designed to implement highly parallel linear algebra algorithms in a serverless setting. We show that, for certain linear algebra algorithms such as matrix multiply, singular value decomposition, and Cholesky decomposition, numpywren's performance (completion time) is within 33% of ScaLAPACK, and its compute efficiency (total CPU-hours) is up to 240% better due to elasticity, while providing an easier to use interface and better fault tolerance. At the same time, we show that the inability of serverless runtimes to exploit locality across the cores in a machine fundamentally limits their network efficiency, which limits performance on other algorithms such as QR factorization. This highlights how cloud providers could better support these types of computations through small changes in their infrastructure.