Scaling in a Distributed Spatial Cache Overlay

This is the source code of our paper

Alexander Gessler, Simon Hanna, Ashley Marie Smith (2014/03) Scaling in a Distributed Spatial Cache Overlay. Paper presented at “Informatiktage 2014: Big Data is Beautiful”, Potsdam. Proceedings in: GI-Edition: Lecture Notes in Informatics: Gesellschaft fuer Informatik (GI) [to appear]

including both the LATEX source of the paper, as well as the software written for it. Please contact the repository owners or file an issue with any questions you may have.

Abstract

Location-based services for mobile devices are a type of distributed system that utilizes geographic behavior of its users. Balancing dynamic query workloads and skewed data remains a problem. Scale-in and scale-out are two proposals that temporarily remove or add resources, respectively. To characterize situations where scaling is more efficient, we implemented a distributed spatial cache overlay for 2D data with the goal of evaluating system performance with and without scaling-out. In this paper, we present an experimental setup to benchmark such a system, and measurements of relative scalability under different cache overlay sizes, query rates and workload distributions. Our results indicate that the system achieves almost linear relative scalability for both uniform and non-uniform query distributions.