Home // ACCSE 2023, The Eighth International Conference on Advances in Computation, Communications and Services // View article

Performance and Scalability of Datastore Technologies for Software Analysis Models

Authors:
Kanishqk Singh
Robert J. Walker

Keywords: Software analysis models; persistence; datastore; performance; scalability

Abstract:
Software Development and Analysis Tools (SDATs) typically contain complex models (software analysis models) that are expensive to compute, and whose expense grows rapidly as the size of the software system under analysis increases. When these models are not stored in a manner that allows them to be restored after program restart, that expense is not amortized; re-computation results in undesirable downtime in the developer’s daily workflow. We investigate options for storing and restoring software analysis models relative to a realistic set of use cases for SDATs. Existing work to study and identify optimal storage technology has been evaluated using datasets either that consist of random graphs—not simulating the nature of real world software—or that derive from excessively small software systems for which recomputing would be feasible. We perform an experimental study on the performance and scalability of datastore technologies exemplifying different approaches (flat files, relational databases, graph databases). We find that SDATs that are heavily focused on storing/retrieving models would find PostgreSQL (a relational database approach) to be the better fit. SDATs that are inclined towards analyzing a limited quantity of software at a given time but involving high maintenance of the models in the database would find Neo4j (a graph database approach) to be the most suitable option.

Pages: 1 to 10

Copyright: Copyright (c) IARIA, 2023

Publication date: June 26, 2023

Published in: conference

ISSN: 2519-8459

ISBN: 978-1-68558-076-6

Location: Nice, France

Dates: from June 26, 2023 to June 30, 2023