TR-08-16: Performance evaluation on GlobeCon – a scalable context management framework

Abstract

Designing a large scale context aggregation and provision system that integrates distributed, heterogeneous and disparate context providers and consumers is a challenging problem. As a system grows numerically or expands geographically, functional building blocks can quickly become bottlenecks in terms of reliability as well as performance. In this paper, we present performance evaluation results of our context management system, GlobeCon, which we have designed recently[19]. GlobeCon is a scalable framework that supports distributed context col- lection, aggregation, processing, provision and usage in large scale ubiq- uitous computing environments. We present a set of evaluation criteria and matrices: i. overhead; ii. effectiveness; iii. adaptability; and iv. scala-bility/throughput, which are used for measuring the overall performance of GlobeCon. We achieved very good results in performance as well as resource utilization. Experimental results show that GlobeCon hierarchical design achieved 100% throughput under the load of 15,000 msg/sec with 1500 concurrent sensors, and 1000 concurrent consumers. Load tests on GlobeCon, which tested the limits of some of the components, show that a LoCoM can achieve 90% throughput under the load of 300,000 msg/sec. Moreover the message delay under extreme load is only tens of milliseconds and remains constant throughout the tests.

TR-08-16.pdf