File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: Latency comparison of replication and coding for data access under random scheduling

TitleLatency comparison of replication and coding for data access under random scheduling
Authors
Issue Date2018
PublisherIEEE. The Journal's web site is located at https://ieeexplore.ieee.org/xpl/conhome.jsp?punumber=1000104
Citation
Proceedings of 2018 IEEE International Conference on Communications (ICC), Kansas City, MO, USA, 20-24 May 2018 How to Cite?
AbstractReplication and coding are two popular approaches to combat failures in large-scale distributed storage systems. Access latency in such systems greatly impacts user experience. Compared with redundant scheduling, random scheduling can reduce the system load, resulting in lower latency, especially when the request arrival rate is high. Besides, random scheduling can achieve near optimal load balancing at a reduced communication cost. A latency comparison of replication and coding is of great importance. Although it has been much argued that coding can achieve lower latency than replication, the latency comparison under random scheduling is still lacking. In this work, based on random scheduling, we analyze the latency of replication and coding and find that, when each request desires all data in a codeword, they have the same average latency. Additionally, we study a general case that users only request a subset of the erasure-coded content, and propose flexible random scheduling for coding, which can lower latency and realize load balancing. Our analysis demonstrates that, in this case, replication achieves lower latency when the system load is low and suffers higher latency when the system load becomes high. With real service time traces from Amazon S3, we conduct trace-driven simulations to validate our analysis.
Persistent Identifierhttp://hdl.handle.net/10722/261957
ISSN

 

DC FieldValueLanguage
dc.contributor.authorShuai, Q-
dc.contributor.authorLi, VOK-
dc.contributor.authorLu, Z-
dc.contributor.authorCao, M-
dc.date.accessioned2018-09-28T04:50:57Z-
dc.date.available2018-09-28T04:50:57Z-
dc.date.issued2018-
dc.identifier.citationProceedings of 2018 IEEE International Conference on Communications (ICC), Kansas City, MO, USA, 20-24 May 2018-
dc.identifier.issn1938-1883-
dc.identifier.urihttp://hdl.handle.net/10722/261957-
dc.description.abstractReplication and coding are two popular approaches to combat failures in large-scale distributed storage systems. Access latency in such systems greatly impacts user experience. Compared with redundant scheduling, random scheduling can reduce the system load, resulting in lower latency, especially when the request arrival rate is high. Besides, random scheduling can achieve near optimal load balancing at a reduced communication cost. A latency comparison of replication and coding is of great importance. Although it has been much argued that coding can achieve lower latency than replication, the latency comparison under random scheduling is still lacking. In this work, based on random scheduling, we analyze the latency of replication and coding and find that, when each request desires all data in a codeword, they have the same average latency. Additionally, we study a general case that users only request a subset of the erasure-coded content, and propose flexible random scheduling for coding, which can lower latency and realize load balancing. Our analysis demonstrates that, in this case, replication achieves lower latency when the system load is low and suffers higher latency when the system load becomes high. With real service time traces from Amazon S3, we conduct trace-driven simulations to validate our analysis.-
dc.languageeng-
dc.publisherIEEE. The Journal's web site is located at https://ieeexplore.ieee.org/xpl/conhome.jsp?punumber=1000104-
dc.relation.ispartofIEEE International Conference on Communications (ICC)-
dc.rightsIEEE International Conference on Communications (ICC). Copyright © IEEE.-
dc.titleLatency comparison of replication and coding for data access under random scheduling-
dc.typeConference_Paper-
dc.identifier.emailLi, VOK: vli@eee.hku.hk-
dc.identifier.emailLu, Z: zhiyilv@hku.hk-
dc.identifier.authorityLi, VOK=rp00150-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1109/ICC.2018.8422492-
dc.identifier.hkuros292175-
dc.publisher.placeUnited States-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats