Journey to MARS: Interplanetary coding for relieving CDNs

Publikation: Beitrag in Buch/Konferenzbericht/Sammelband/GutachtenBeitrag in KonferenzbandBeigetragen

Abstract

The amount of consumer data transmitted over the internet will increase in the future. At the moment, this traffic is mainly handled by Content Delivery Networks (CDN) via Hypertext Transfer Protocol (HTTP). However, this server-based approach has the disadvantage that the servers themselves and their connections to the Internet are under a high load, which makes them a bottleneck. Therefore, new approaches for efficient content distribution are sought. One outstanding approach is the Interplanetary File System (IPFS), the synthesis of various successful Peer-to-Peer (P2P) approaches. Theoretically, the load on the server can be reduced by distributing the data to several nodes. To transmit data optimally, however, a high degree of coordination between nodes is necessary. Research on other content distribution schemes has shown that the cooperation effort can be avoided by using Random Linear Network Coding (RLNC). This has not been used in IPFS so far. In this work we present Multi Access Recoding System (MARS), a protocol which combines IPFS with RLNC. Simulations on the Interplanetary Testbed (IPTB) have shown that MARS is able to reduce the server load and download time by up to 50% compared to a transmission using only a single server and up to 45% compared to the same setup without RLNC. Even without coordination, this only causes an additional network load of 30%. In addition, MARS improves the download time by at least 15% even for very few peer nodes or peers with little storage.

Details

OriginalspracheUndefiniert
TitelGLOBECOM 2020 - 2020 IEEE Global Communications Conference
Seiten1-6
ISBN (elektronisch)978-1-7281-8298-8
PublikationsstatusVeröffentlicht - 1 Dez. 2020
Peer-Review-StatusNein

Publikationsreihe

ReiheIEEE Conference on Global Communications (GLOBECOM)
ISSN1930-529X

Externe IDs

Scopus 85100382075
ORCID /0000-0001-8469-9573/work/161891015

Schlagworte