Reliable publish/subscribe in content-centric networks

  • Authors:
  • Jiachen Chen;Mayutan Arumaithurai;Xiaoming Fu;K.K. Ramakrishnan

  • Affiliations:
  • Institute of Computer Science, University of Goettingen, Goettingen, Germany;NEC Laboratories Europe, Heidelberg, Germany;Institute of Computer Science, University of Goettingen, Goettingen, Germany;AT&T Labs-Research, Florham Park, NJ, USA

  • Venue:
  • Proceedings of the 3rd ACM SIGCOMM workshop on Information-centric networking
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

Managing congestion is a challenge in content-centric networks due to the lack of an end-to-end session context over which a `flow' may be controlled. Flow and congestion control as well as reliable delivery are often considered even more of a challenge in content-centric publish/subscribe systems, where the nature of information dissemination is similar to multicast. These have been long-standing challenges also for IP multicast. With an unknown number of publishers, a content-centric pub/sub environment exacerbates these problems, demanding new solutions. In this paper, we propose a lightweight enhancement to content-centric publish/subscribe systems for flow and congestion control as well as for reliability. R-COPSS allows the publishers to efficiently use the content-centric network by having subscribers generate timely feedback while enabling subscribers to make use of NDN to perform local repair for reliable delivery. Rather than having all subscribers generate feedback (ACK) per packet, we seek to elect particular subscribers in a hierarchy to provide the feedback and the rest of them resort to a periodical summary. Our approach not only reduces the load on the publisher, but also removes the requirement on the publisher to limit its sending rate to the slowest subscriber. Our preliminary results show that R-COPSS performs better in terms of overall throughput and is fair to competing flows.