Interval graph multi-coloring-based resource reservation for energy-efficient containerized cloud data centers
- PDF / 5,998,817 Bytes
- 49 Pages / 439.37 x 666.142 pts Page_size
- 97 Downloads / 184 Views
Interval graph multi‑coloring‑based resource reservation for energy‑efficient containerized cloud data centers Yashwant Singh Patel1 · Anjali Baheti2 · Rajiv Misra1 Accepted: 18 September 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Containerized deployment of microservices has quickly become a well-known virtualization technology due to its higher portability, scalability, good isolation, and lightweight solutions. However, it faces several challenges in terms of the capital and operational expenses in large-scale data centers. In particular, services in cloud are usually instantiated as a group of containers, which continuously trigger frequent communication workloads and hence significantly degrades the service performance due to inefficient allocation of containers. Thus to deploy microservices, service providers must consider different types of objectives, such as optimizing the communication cost or the operational cost, which are joint objectives that have previously been studied independently. In this paper, we study the problem of communicationaware container-based advance reservation to optimize the energy and communication cost for microservices deployment. We applied the interval graph model to map the container reservation scenario of microservices and derived various performance bounds. Then, we propose greedy graph multi-coloring-based centralized and distributed algorithms to find an efficient allocation of containers. Through theoretical analysis and extensive experimental results, we demonstrate that the proposed approaches can decrease the total cost by up to 31% compared to the current stateof-the-art methods. Keywords Containerization · Cloud data centers · Interval graph · Container communication · Energy-efficient
* Yashwant Singh Patel [email protected] Anjali Baheti [email protected] Rajiv Misra [email protected] 1
Department of Computer Science and Engineering, Indian Institute of Technology Patna, Patna, India
2
Department of Computer Science and Engineering, Medi-Caps University, Indore, India
13
Vol.:(0123456789)
Y. S. Patel et al.
1 Introduction Modern cloud deployments massively rely on hypervisor-based virtualization. However, the ubiquitous features of Docker and Container technology have given them momentum in their extensive selection as alternatives for shaping the future of software development. Since containers enable lightweight, highly-portable, scalable properties and offer performance isolation, developers are accepting the containerization technique, and enterprises are adopting it at an explosive rate in addition to hardware virtualization [1–3]. Recently, the development of software containerization tools such as LXC [4] and Docker [5] enables cloud users to build containers and run various kinds of applications on virtual clusters in cloud data centers. To efficiently utilize and manage the container resources in a clustered environment, cloud service providers prefer a shared cluster to deploy thei
Data Loading...
