Rich interactive applications on the Internet require low latency communication in order to deliver a responsive user experience. This problem is compounded when multiple users interact in an application where the user with the highest delay (i.e., lag) causes a drop in quality of experience (QoE) for the rest of the group. Large organizations, with the luxury of private infrastructure, can place this infrastructure (e.g., proxies, VPNs) close to the end user in order to mitigate high network latencies. However, smaller-scale applications do not have this ability. In this paper, the authors consider the challenge of reducing latency for small-scale Internet applications, namely those without the ability to place data centers/physical resources near to the users. In this context, the authors explore challenges of deploying Internet services across multiple cloud providers. They show that, for small-scale applications, it is possible to trade off latency and cost, even in multi-cloud deployments. They further expand their analysis to consider provisioning for interactive group applications. In this setting, the authors formulate the "Cooperative Group Provisioning" (CGP) problem which aims to bound latency and minimize costs for group applications. The authors prove that it is NP-hard to even approximate within a constant factor the optimal solution to CGP and present and evaluate a greedy algorithm to solve the problem. The reviewers agree that the problem tack-led in this paper is an important one, especially given the rise of collaborative online applications. It also raises interesting questions for future work about how to adapt the proposed greedy algorithm to a distributed or online setting.
展开▼
