Improving Cloud Service Resilience Using Brownout-Aware Load-Balancing

Abstract

We focus on improving resilience of cloud services (e.g., e-commerce website), when correlated or cascading failures lead to computing capacity shortage. We study how to extend the classical cloud service architecture composed of a loadbalancer and replicas with a recently proposed self-adaptive paradigm called brownout. Such services are able to reduce their capacity requirements by degrading user experience (e.g., disabling recommendations).Combining resilience with the brownout paradigm is to date an open practical problem. The issue is to ensure that replica self-adaptivity would not confuse the load-balancing algorithm, overloading replicas that are already struggling with capacity shortage. For example, load-balancing strategies based on response times are not able to decide which replicas should be selected, since the response times are already controlled by the brownout paradigm.In this paper we propose two novel brownout-aware loadbalancing algorithms. To test their practical applicability, we extended the popular lighttpd web server and load-balancer, thus obtaining a production-ready implementation. Experimental evaluation shows that the approach enables cloud services to remain responsive despite cascading failures. Moreover, when compared to Shortest Queue First (SQF), believed to be nearoptimal in the non-adaptive case, our algorithms improve user experience by 5%, with high statistical significance, while preserving response time predictability.

Event-Driven Application Brownout: Reconciling High Utilization and Low Tail Response Times

“…As the initial periodic brownout has been tested with load balancing algorithms [5], the event-based version of Brownout could also be tested with these algorithms as part of a future work. Both Brownout versions avoid overloads, but the eventbased version described in this paper also obtains response times closer to the desired set-point.…”

“…However, we henceforth consider only one VM as the focus of this paper is on the performance of our eventdriven Brownout algorithms. Part of an envisaged future work is to include the event-driven Brownout algorithms with load balancing algorithms, just as it was done with the initial periodic Brownout [5]. Figure 1 represents the deployment of Brownout in a proxy interacting with a single VM.…”

“…This is mostly due to two reasons: the periodic nature of the controller and the fact that controller decisions are based on response time alone. Experience with the related problem of load balancing [5] suggests that improvements can be achieved by taking a control decision for each request, i.e., in an eventdriven manner, and by basing these decisions also on queuelength.…”

Energy‐aware cross‐layer resource allocation in mobile cloud

“…Gulshan Soni et al proposed an efficient load balancing policy based on a central load balancer to balance the load among virtual machines in a cloud data center. Cristian Klein et al presented a novel approach for improving resilience of cloud services by using brownout‐aware load balancing. Ankita Taneja et al proposed a proficient metaheuristic cloud load balancing algorithm, which uses ant colony algorithm to tackle the balancing of burden among the cloud nodes.…”

Software Engineering Meets Control Theory

“…One common way to validate a controller implementation and its system under control is to show statistical evidence, for example, using cumulative distribution functions, as done in [42,43]. In this work, less experiment effort is required are than for the rigorous analysis mentioned previously.…”