Abstract
Traditionally, networks such as datacenter interconnects are designed to optimize worstcase performance under arbitrary traffic patterns. Such network designs can however be far from optimal when considering the actual workloads and traffic patterns which they serve. This insight led to the development of demandaware datacenter interconnects which can be reconfigured depending on the workload.
Motivated by these trends, this paper initiates the algorithmic study of demandaware networks (DANs), and in particular the design of boundeddegree networks. The inputs to the network design problem are a discrete communication request distribution, D, defined over communicating pairs from the node set V, and a bound, d, on the maximum degree. In turn, our objective is to design an (undirected) demandaware network N = (V,E) of boundeddegree d, which provides short routing paths between frequently communicating nodes distributed across N. In particular, the designed network should minimize the expected path length on N (with respect to D), which is a basic measure of the efficiency of the network.
We show that this fundamental network design problem exhibits interesting connections to several classic combinatorial problems and to information theory. We derive a general lower bound based on the entropy of the communication pattern D, and present asymptotically optimal networkaware design algorithms for important distribution families, such as sparse distributions and distributions of locally bounded doubling dimensions.
BibTeX  Entry
@InProceedings{avin_et_al:LIPIcs:2017:8015,
author = {Chen Avin and Kaushik Mondal and Stefan Schmid},
title = {{DemandAware Network Designs of Bounded Degree}},
booktitle = {31st International Symposium on Distributed Computing (DISC 2017)},
pages = {5:15:16},
series = {Leibniz International Proceedings in Informatics (LIPIcs)},
ISBN = {9783959770538},
ISSN = {18688969},
year = {2017},
volume = {91},
editor = {Andr{\'e}a W. Richa},
publisher = {Schloss DagstuhlLeibnizZentrum fuer Informatik},
address = {Dagstuhl, Germany},
URL = {http://drops.dagstuhl.de/opus/volltexte/2017/8015},
URN = {urn:nbn:de:0030drops80153},
doi = {10.4230/LIPIcs.DISC.2017.5},
annote = {Keywords: Network design, reconfigurable networks, datacenter topology, peertopeer computing, entropy, sparse spanners}
}
Keywords: 

Network design, reconfigurable networks, datacenter topology, peertopeer computing, entropy, sparse spanners 
Seminar: 

31st International Symposium on Distributed Computing (DISC 2017) 
Issue Date: 

2017 
Date of publication: 

05.10.2017 