Problems in Modern High Performance Parallel I/O Systems

TL;DR

This paper discusses the challenges faced by modern high-performance parallel I/O systems, highlighting bottlenecks, suboptimal data practices, and future optimization opportunities in the context of rapidly scaling supercomputers.

Contribution

It provides an analysis of current I/O bottlenecks, examines data handling practices, and explores potential future improvements in high-performance I/O systems.

Findings

I/O bottlenecks are a major performance bottleneck in supercomputers.

Scientific applications often do not utilize peak I/O bandwidth.

Current data practices may need reevaluation for future scalability.

Abstract

In the past couple of decades, the computational abilities of supercomput- ers have increased tremendously. Leadership scale supercomputers now are capable of petaflops. Likewise, the problem size targeted by applications running on such computers has also scaled. These large applications have I/O throughput requirements on the order of tens of gigabytes per second. For a variety of reasons, the I/O subsystems of such computers have not kept pace with the computational increases, and the time required for I/O in an application has become one of the dominant bottlenecks. Also troublesome is the fact that scientific applications do not attain near the peak theoretical bandwidth of the I/O subsystems. In addressing the two prior issues, one must also question the nature of the data itself; one can ask whether contem- porary practices of data dumping and analysis are optimal and whether they will continue to be applicable as computers continue to scale. These three topics, the I/O subsystem, the nature of scientific data output, and future possible optimizations are discussed in this report.