Zobrazeno 1 - 7
of 7
pro vyhledávání: '"Bairen, Yi"'
Long tail latency of short flows (or messages) greatly affects user-facing applications in datacenters. Prior solutions to the problem introduce significant implementation complexities, such as global state monitoring, complex network control, or non
Externí odkaz:
http://arxiv.org/abs/2110.15157
Publikováno v:
IEEE INFOCOM 2022 - IEEE Conference on Computer Communications.
Autor:
Chuanxiong Guo, Chang Lan, Yibo Zhu, Yanghua Peng, Yixin Bao, Bairen Yi, Chuan Wu, Yangrui Chen
Publikováno v:
SOSP
We present ByteScheduler, a generic communication scheduler for distributed DNN training acceleration. ByteScheduler is based on our principled analysis that partitioning and rearranging the tensor transmissions can result in optimal results in theor
Publikováno v:
SIGCOMM Posters and Demos
Remote Direct Memory Access (RDMA) offers ultra-low latency and CPU bypass networking to application programmers. Existing applications are often designed around socket based software stack that manages application buffers separately from networking
Autor:
Danny Price, Ellert van der Velden, Sébastien Celles, Pieter Eendebak, Michael McKerns, Eben Olson, Colin Raffel, Bairen Yi, Elliott Ash
Publikováno v:
Journal of Open Source Software. 3:1115
Publikováno v:
SIGCOMM
Leveraging application-level requirements using coflows has recently been shown to improve application-level communication performance in data-parallel clusters. However, existing coflow-based solutions rely on modifying applications to extract coflo