Resource Utilization Monitoring for Raw Data Query Processing

TL;DR

This paper enhances raw data query processing by integrating resource monitoring, analyzing resource usage patterns, and proposing adaptive data partitioning techniques to improve efficiency and scalability in scientific data analysis.

Contribution

It introduces a resource monitoring module into raw data query frameworks and proposes QCA and RUA data partitioning techniques based on resource usage analysis.

Findings

Sampling queries have the lowest resource utilization.

PostgresRAW outperforms PostgreSQL on simple 0-JOIN queries.

Complex JOIN queries benefit from PostgreSQL to reduce workload.

Abstract

Scientific experiments, simulations, and modern applications generate large amounts of data. Data is stored in raw format to avoid the high loading time of traditional database management systems. Researchers have proposed many techniques to improve query execution time for raw data and reduce data loading time for traditional systems. The core of all the proposed techniques is efficient utilization of resources by processing only required data or reducing operations on data. The processed data caching in the main memory or disk can resolve this issue and avoid repeated processing of data. However, limitations of resources like main memory space, storage IO speeds, and additional storage space requirements on disk need to be considered to provide reliable and scalable solutions for cloud or in-house deployments. This paper presents improvements to the raw data query processing framework by integrating a resource monitoring module. The experiments were performed using a scientific dataset known Sloan Digital Sky Survey (SDSS). Analysis of monitored resources revealed that sampling queries had the lowest resource utilization. The PostgresRAW can answer simple 0-JOIN queries faster than PostgreSQL. While one or more JOIN complex queries need to be answered using PostgreSQL to reduce workload execution time (WET). The results section discusses resource requirements of simple, complex, and sampling type queries. The result analysis of query types and resource utilization patterns assisted in proposing Query Complexity Aware (QCA) and Resource Utilization Aware (RUA) data partitioning techniques for raw engines and DBMS to reduce cost or data to result time.