Cloud computing is a popular way to address the scalability and efficiency issues of data centers. While the level of development of cloud technologies is already high-enough to easily beat old static cluster configurations, there is still a lot of room for improvement. One of these areas is in the way resource management tools predict the CPU and tasks performance.Normally, resource managers assume that the tasks do not afTect each other, and assign resources under this assumption. With a simple experiment this paper shows that this assumption is grossly wrong, leading to overestimation of task performance that can approach 50%. Next, the paper presents a behavioral model that efficiently addresses these issues. Preliminary results obtained for three different hardware platforms demonstrate the benefits of our model in performance prediction.
Abstract-Virtualization is a key enabler technology for cloud computing. It allows applications to share computing, memory, storage, and network resources. However, physical resources are not standalone and the server infrastructure is not homogeneous. The CPU cores are commonly connected to the shared memory, caches, and computational units. As a result, the performance of cloud applications can be greatly affected if, while being executed at different computing cores, they compete for the same shared cache or network resource. The performance degradation can be as high as 50%. In this work we present a methodology which predicts the performance problems of cloud applications during their concurrent execution by looking at the hardware performance counters collected during their standalone execution. The proposed methodology fosters design of novel solutions for efficient resource allocation and scheduling.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.