Load balancing in Unihub SaaS system based on user behavior prediction
https://doi.org/10.15514/ISPRAS-2015-27(5)-2
Abstract
In ISP RAS cloud computing system SaaS Unihub was developed. It provides the possibility for users to work by Web-browser with interactive graphic Linux-applications, working in isolated Docker containers. Containers have dynamical demands to computational resources. Usual way of placement when containers are distributed uniformly among all servers can lead to bad result: some servers have too many applications but other are almost empty. In this paper we propose to collect information about users behavior and investigate how different applications work in order to predict containers load-balancing. Our observations show that such information can provide more uniform load-balancing and improve the whole system performance.
About the Authors
D. A. Grushin
ISP RAS
Russian Federation
Russian Federation
N. N. Kuzyurin
ISP RAS
Russian Federation
Russian Federation
References
1. Atkeson C., Moore A., Schaal S. Locally weighted learning. Artificial Intelligence Review. 1997. № 1-5 (11). C. 11–73.
2. Blum A. On-line algorithms in machine learning Springer, 1996. 306–325 с.
3. Deboosere L. [и др.]. Cloud-based desktop services for thin clients. Internet Com-puting, IEEE. 2012. № 6 (16). C. 60–67.
4. Jiang Y. [и др.]. ASAP: A self-adaptive prediction system for instant cloud re-source demand provisioning 2011. 1104–1109 с.
5. Kavulya S. [и др.]. An analysis of traces from a production mapReduce cluster 2010.C. 94–103.
6. Nguyen T.-D. [и др.]. Prediction-based energy policy for mobile virtual desktop infrastructure in a cloud environment. Inf. Sci. 2015. № C (319). C. 132–151.
7. Smith W. Prediction services for distributed computing 2007. C. 1–10.
8. Suznjevic M., Skorin-Kapov L., Humar I. User behavior detection based on statis-tical traffic analysis for thin client services Advances in intelligent systems and computing. Springer International Publishing, 2014. C. 247–256.
9. Wilson D.R., Martinez T.R. Improved heterogeneous distance functions. J. Artif. Int. Res. 1997. № 1 (6). C. 1–34.
10. Unihub: Technological platform for university cluster program. http://unihub.ru.
11. Docker: An open platform for distributed applications for developers and sysad-mins. http://www.docker.com.
12. CRIU: A project to implement checkpoint/restore functionality for linux in us-erspace. http://criu.org/Docker.
13. Mohammadi F., Jamali S., Bekvari M., Survey on Job Scheduling algorithms in Cloud Computing, Int. Journal of Emergency Trends of Technology in Computer Science, 2014, v. 3, Issue 2, C. 151-154.
14. Tian W., Zhao E., Zhong V., Xu M., Jing C., A dynamic and integrated load-balancing scheduling algorithm for cloud datacenters, Cloud computing and Intel-ligence Systems (CCIS), 2011, IEEE Int. Conference15-17 September 2011, C. 311-315.
15. Duy T.V.T., Sato Y., Inogushi Y., Performance evaluation of a green scheduling algorithm for energy savings in cloud computing, Parallel and Distributed Pro-cessing, Workshops and PhD forum, Atlanta, 19-23 April 2010, C. 1-8.
Review
For citations:
Grushin D.A., Kuzyurin N.N. Load balancing in Unihub SaaS system based on user behavior prediction. Proceedings of the Institute for System Programming of the RAS (Proceedings of ISP RAS). 2015;27(5):23-34. (In Russ.) https://doi.org/10.15514/ISPRAS-2015-27(5)-2
Email this article 