Improving a High Productivity Data Analytics Chapel Framework

TL;DR

This paper enhances the Arkouda data analytics framework by introducing client-side optimizations that significantly boost performance while preserving its interactive capabilities, through techniques like operation delaying, caching, and subexpression elimination.

Contribution

The paper proposes novel client-side optimization techniques for Arkouda that improve performance without sacrificing interactivity, including operation delaying, caching, and result reuse.

Findings

Performance improvements of 20% to 120% across benchmarks

Maintains full interactivity of the framework

Effective caching and operation reuse strategies

Abstract

Most state of the art exploratory data analysis frameworks fall into one of the two extremes: they either focus on the high-performance computational, or on the interactive and open-ended aspects of the analysis. Arkouda is a framework that attempts to integrate the interactive approach with the high performance computation by using a novel client-server architecture, with a Python interpreter on the client side for the interactions with the scientist and a Chapel server for performing the demanding high-performance computations. The Arkouda Python interpreter overloads the Python operators and transforms them into messages to the Chapel server that performs the actual computation. In this paper, we are proposing several client-side optimization techniques for the Arkouda framework that maintain the interactive nature of the Arkouda framework, but at the same time significantly improve the performance of the programs that perform operations running on the high-performance Chapel server. We do this by intercepting the Python operations in the interpreter, and delaying their execution until the user requires the data, or we fill out the instruction buffer. We implement caching and reuse of the Arkouda arrays on the Chapel server side (thus saving on the allocation, initialization and deallocation of the Chapel arrays), tracking and caching the results of function calls on the Arkouda arrays (thus avoiding repeated computation) and reusing the results of array operations by performing common subexpression elimination. We evaluate our approach on several Arkouda benchmarks and a large collection of real-world and synthetic data inputs and show significant performance improvements between 20% and 120% across the board, while fully maintaining the interactive nature of the Arkouda framework.