Skip to main navigation Skip to search Skip to main content

ABENA: An Ageing before Temperature Electromigration-Aware Neighbour Allocation for Many-Core Architectures

    Research output: Contribution to Book/ReportConference Contributionpeer-review

    Abstract

    The percentage of inactive nodes or dark nodes (we refer to dark nodes as dumso) in many-core systems increases because of power dissipation caused by continuous scaling in technology. To address this challenge, existing work employ several techniques. Some techniques place the dumso nodes strategically between active nodes to alleviate the temperature by choosing the nodes which are far away from each other. However, this increases the latency between nodes which require inter communication leading to performance degradation. Others employ Dynamic Thermal Management (DTM) to vary the Voltage Frequency (V/F) Scaling of the nodes whilst Task migration is used to migrate tasks. In this paper, an Ageing Before Temperature Eletromigration-Aware Neighbour Allocation (ABENA) is proposed to alleviate the temperature of the nodes by using the Lifetime of nodes as the main parameter. Experiments show that our approach improves the lifetime of nodes.
    Original languageEnglish
    Title of host publication4th International Conference on Data Storage and Data Engineering
    PublisherAssociation for Computing Machinery (ACM)
    ISBN (Print)978-1-4503-8930-3
    DOIs
    Publication statusPublished - 18 Feb 2021
    EventDSDE '21: 4th International Conference on Data Storage and Data Engineering -
    Duration: 18 Feb 202120 Feb 2021
    Conference number: 4th

    Publication series

    Name2021 4th International Conference on Data Storage and Data Engineering

    Conference

    ConferenceDSDE '21
    Period18/02/2120/02/21

    Keywords

    • Many-Core
    • Dark-Silicon
    • Embedded Systems
    • Network-on-Chip
    • NoC

    Fingerprint

    Dive into the research topics of 'ABENA: An Ageing before Temperature Electromigration-Aware Neighbour Allocation for Many-Core Architectures'. Together they form a unique fingerprint.

    Cite this