TRINI: an adaptive load balancing strategy based on garbage collection for clustered Java systems

Abstract

Summary Nowadays, clustered environments are commonly used in high‐performance computing and enterprise‐level applications to achieve faster response time and higher throughput than single machine environments. Nevertheless, how to effectively manage the workloads in these clusters has become a new challenge. As a load balancer is typically used to distribute the workload among the cluster's nodes, multiple research efforts have concentrated on enhancing the capabilities of load balancers. Our previous work presented a novel adaptive load balancing strategy (TRINI) that improves the performance of a clustered Java system by avoiding the performance impacts of major garbage collection, which is an important cause of performance degradation in Java. The aim of this paper is to strengthen the validation of TRINI by extending its experimental evaluation in terms of generality, scalability and reliability. Our results have shown that TRINI can achieve significant performance improvements, as well as a consistent behaviour, when it is applied to a set of commonly used load balancing algorithms, demonstrating its generality. TRINI also proved to be scalable across different cluster sizes, as its performance improvements did not noticeably degrade when increasing the cluster size. Finally, TRINI exhibited reliable behaviour over extended time periods, introducing only a small overhead to the cluster in such conditions. These results offer practitioners a valuable reference regarding the benefits that a load balancing strategy, based on garbage collection, can bring to a clustered Java system. Copyright © 2016 John Wiley & Sons, Ltd.

COCOA: A Synthetic Data Generator for Testing Anonymization Techniques

“…In terms of costs, our main metrics were execution time, CPU (%) and memory (MB) utilizations. Garbage collection (GC) was also monitored as it is an important performance concern in Java [25].…”

Towards an emulated IoT test environment for anomaly detection using NEMU

“…The sensor temperature data was simulated by using a script which randomly generated temperature values by following a normal distribution centered at 20 C and a 2σ of 0.2 C. It ran on the devices found on the leaf nodes of the network, transmitting this information along with the other metrics. To generate processor and memory loads, we introduced some controlled noise to the system by running on the devices a subset of the widely-used Java Dacapo benchmarks [12].…”

“…Even though such scenarios might not be the most commonly found in IoT (as the data gathering and processing usually tend to be light-weight in terms of CPU and memory), the aim was to strengthen the validation of our solution by identifying either other potentially applicable usage scenarios or limitations on its usage. The experimental set-up was similar to that used in the previous experiment, with the following difference: instead of using MQTT, the tested devices ran a subset of the widely-used Java Dacapo Benchmarks [12]. This strategy allowed us to diversify the assessed behaviours (i.e., CPU and memory usage) by introducing some variability on the workloads processed by the system.…”

In-Test Adaptation of Workload in Enterprise Application Performance Testing

“…Finally, the performance bugs were retrieved from WAIT's outputs, which was fed with Javacores [4] (snapshots of the JVM state). Finally, we built our prototype on top of the JMeter tool with Java, which is an object-oriented programming language widely used for being open source and highly portable [16]. Internally, the prototype supports a first heuristic version of our approach, in which any adjustments to the test workload affects all transaction types.…”