Cloud Application Predictability through Integrated Load-Balancing and Service Time Control

Abstract

Cloud computing provides the illusion of infinite capacity to application developers. However, data center provisioning is complex and it is still necessary to handle the risk of capacity shortages. To handle capacity shortages, graceful degradation techniques sacrifice user experience for predictability. In all these cases, the decision making policy that determines the degradation interferes with other decisions happening at the infrastructure level, like load-balancing choices. Here, we reconcile the two approaches, developing a load-balancing strategy that also handles capacity shortages and graceful degradation when necessary. The proposal is based on a sound control-theoretical approach. The design of the approach avoids the pitfalls of interfering control decisions. We describe the technique and provide evidence that it allows us to achieve higher performance in terms of emergency management and user experience.

Quality-Elasticity: Improved Resource Utilization, Throughput, and Response Times Via Adjusting Output Quality to Current Operating Conditions

“…Various algorithms and controllers have been employed to extend the brownout-concept as originally envisioned, including how it can be used to optimize both service delivery [9], [10], [11], [12], [13] and cloud infrastructure power consumption [14], [15], [16]. These results validate the claim that great utility and efficiency increases can be made using a qualityreduction approach.…”

“…We would also then like to implement and compare with more recent work on Brownout, e.g. [9], [10], [11], [12], [13]. • Study the relationship between utilization and utility, as a measure of how useful customers find a service.…”

Improving Predictability of User-Affecting Metrics to Support Anomaly Detection in Cloud Services

“…As topics for future research, we envision the application of the proposed framework to large mission-critical systems. Specifically, additional experiments are required to evaluate how to integrate admission control and load balancing [2] with the cloud computing infrastructure configuration approach to ensure mean response time predictability. In addition, we plan to compare the performance of anomaly detection algorithms based on different approaches, such as bucket algorithms [53] which can naturally leverage the statistical characterization of response times proposed in this paper against other machine learning algorithms [21].…”

“…The multi-server paradigm with a central queueing system, e.g., encompassing a load balancing front end towards a cloud computing environment, has been used as the state of the practice for scalable server-side computation for the last several years [1], [2], [3]. In spite of the widespread adoption of the multi-server paradigm, the setup of such systems still poses a number of challenges related to the shaping of the workload and the configuration of the servers.…”

“…deviation of response time 95.20 ms 83.72 msNonetheless, the framework is general, and can encompass both endogenous and exogenous attack vectors[30]. Future work consists in extending the experiments to account for a wider set of attack vectors 2.…”