Towards an Energy Efficient Internet Core with Near-Zero Buffers

Speaker:  Arun Vishwanath – Melbourne, VIC, Australia
Topic(s):  Computational Theory, Algorithms and Mathematics

Abstract

Reducing the power consumption of core Internet routers is important for both Internet Service Providers (ISPs) and router vendors. ISPs can reduce their Carbon footprint and operational costs, while router manufacturers can achieve higher switching capacity per rack space. In this lecture we focus on how core routers and networks can be designed with near-zero buffers to enable an energy-efficient Internet.

Part 1

We examine the impact of packet buffers on the power consumption of backbone router line-cards. We argue that Gigabytes of always-on SRAM and DRAM buffers account for around 10% of the power, but are actively used only during transient periods of congestion. We propose a simple and practical algorithm for activating buffers incrementally as needed and putting them to sleep when not in use. We evaluate our algorithm on traffic traces from carrier and enterprise networks, via simulations in ns2, and by implementing it on a programmable router test-bed. Our study shows that much of the energy associated with off-chip packet buffers can be safely eliminated with negligible impact on traffic performance. Dynamic adjustment of active router buffer size provides a low-complexity low-risk mechanism of saving energy that is amenable for incremental deployment in networks today.  

Part 2

Since we propose that buffer size be adjusted at run-time rather than build-time, ISPs can gradually become comfortable with the idea of operating with reduced active buffers, making them more likely to adopt routers built with smaller buffers in the future. Towards this goal, we envision a bufferless (or near-zero buffer) core network and propose a novel edge-to-edge based packet-level forward error correction (FEC) scheme as a means for combating packet loss in the bufferless core. We characterise the impact of FEC strength on loss at a single link and then develop a global optimisation framework for multi-hop networks. We propose a heuristic algorithm that adjusts FEC strength to achieve fairness amongst the different single- and multi-hop flows. Finally, we evaluate the performance of our FEC scheme for realistic mixes of short- and long-lived TCP flows, and show that edge-to-edge packet-level FEC can be tuned to effectively mitigate contention losses in the Internet core, thus opening the doors to a bufferless optical network in the near future.

About this Lecture

Number of Slides:  45
Duration:  90 minutes
Languages Available:  English
Last Updated: 

Request this Lecture

To request this particular lecture, please complete this online form.

Request a Tour

To request a tour with this speaker, please complete this online form.

All requests will be sent to ACM headquarters for review.