A Container Service Chain Placement Greedy Algorithm Based on Heuristic Information

Abstract

Container technology is a light weight back-end virtualization solution which copes with the growing demand for Internet service concurrency. Network services can be provided by a series of containers between which network requests are sent and responded. We call these cooperating containers a container service chain. On a certain physical cluster, there are various available container placement of a service chain. Different container placement causes different service response time. In order to reduce the response time of container service chain, a greedy algorithm for container service chain placement is proposed based on heuristic information. Through quantitative analyzing resources provided by hosts and consumed by containers, we build a mathematic model of container service chain. We use this model to decide where to place container when service is expanded, and where to remove container when service is reduced. Our model reduced the response time of container service chains by 40% compared to average placement.