Matches in SemOpenAlex for { <https://semopenalex.org/work/W3093671646> ?p ?o ?g. }
Showing items 1 to 91 of
91
with 100 items per page.
- W3093671646 abstract "Distributed deep learning (DDL) uses a cluster of servers to train models in parallel. This has been applied to a multiplicity of problems, e.g. online advertisement, friend recommendations. However, the distribution of training means that the communication network becomes a key component in system performance. In this paper, we measure the Alibaba's DDL system, with a focus on understanding the bottlenecks introduced by the network. Our key finding is that the communications overhead has a surprisingly large impact on performance. To explore this, we analyse latency logs of 1.38M Remote Procedure Calls between servers during model training for two real applications of high-dimensional sparse data. We reveal the major contributors of the latency, including concurrent write/read operations of different connections and network connection management. We further observe a skewed distribution of update frequency for individual parameters, motivating us to propose using in-network computation capacity to offload server tasks." @default.
- W3093671646 created "2020-10-29" @default.
- W3093671646 creator A5008795836 @default.
- W3093671646 creator A5023313904 @default.
- W3093671646 creator A5029024939 @default.
- W3093671646 creator A5030611214 @default.
- W3093671646 creator A5030689390 @default.
- W3093671646 creator A5058593618 @default.
- W3093671646 creator A5070412038 @default.
- W3093671646 creator A5085106892 @default.
- W3093671646 date "2020-10-27" @default.
- W3093671646 modified "2023-09-27" @default.
- W3093671646 title "Dissecting the Communication Latency in Distributed Deep Sparse Learning" @default.
- W3093671646 cites W1825216778 @default.
- W3093671646 cites W2018378172 @default.
- W3093671646 cites W2149939304 @default.
- W3093671646 cites W2769986458 @default.
- W3093671646 cites W2931075014 @default.
- W3093671646 cites W2949161920 @default.
- W3093671646 cites W2975712713 @default.
- W3093671646 cites W2996471668 @default.
- W3093671646 cites W4245217724 @default.
- W3093671646 doi "https://doi.org/10.1145/3419394.3423637" @default.
- W3093671646 hasPublicationYear "2020" @default.
- W3093671646 type Work @default.
- W3093671646 sameAs 3093671646 @default.
- W3093671646 citedByCount "5" @default.
- W3093671646 countsByYear W30936716462021 @default.
- W3093671646 countsByYear W30936716462022 @default.
- W3093671646 crossrefType "proceedings-article" @default.
- W3093671646 hasAuthorship W3093671646A5008795836 @default.
- W3093671646 hasAuthorship W3093671646A5023313904 @default.
- W3093671646 hasAuthorship W3093671646A5029024939 @default.
- W3093671646 hasAuthorship W3093671646A5030611214 @default.
- W3093671646 hasAuthorship W3093671646A5030689390 @default.
- W3093671646 hasAuthorship W3093671646A5058593618 @default.
- W3093671646 hasAuthorship W3093671646A5070412038 @default.
- W3093671646 hasAuthorship W3093671646A5085106892 @default.
- W3093671646 hasConcept C108583219 @default.
- W3093671646 hasConcept C111919701 @default.
- W3093671646 hasConcept C11413529 @default.
- W3093671646 hasConcept C120314980 @default.
- W3093671646 hasConcept C120665830 @default.
- W3093671646 hasConcept C121332964 @default.
- W3093671646 hasConcept C154945302 @default.
- W3093671646 hasConcept C172086080 @default.
- W3093671646 hasConcept C192209626 @default.
- W3093671646 hasConcept C26517878 @default.
- W3093671646 hasConcept C2779960059 @default.
- W3093671646 hasConcept C29140674 @default.
- W3093671646 hasConcept C31258907 @default.
- W3093671646 hasConcept C41008148 @default.
- W3093671646 hasConcept C45374587 @default.
- W3093671646 hasConcept C76155785 @default.
- W3093671646 hasConcept C82876162 @default.
- W3093671646 hasConcept C93996380 @default.
- W3093671646 hasConceptScore W3093671646C108583219 @default.
- W3093671646 hasConceptScore W3093671646C111919701 @default.
- W3093671646 hasConceptScore W3093671646C11413529 @default.
- W3093671646 hasConceptScore W3093671646C120314980 @default.
- W3093671646 hasConceptScore W3093671646C120665830 @default.
- W3093671646 hasConceptScore W3093671646C121332964 @default.
- W3093671646 hasConceptScore W3093671646C154945302 @default.
- W3093671646 hasConceptScore W3093671646C172086080 @default.
- W3093671646 hasConceptScore W3093671646C192209626 @default.
- W3093671646 hasConceptScore W3093671646C26517878 @default.
- W3093671646 hasConceptScore W3093671646C2779960059 @default.
- W3093671646 hasConceptScore W3093671646C29140674 @default.
- W3093671646 hasConceptScore W3093671646C31258907 @default.
- W3093671646 hasConceptScore W3093671646C41008148 @default.
- W3093671646 hasConceptScore W3093671646C45374587 @default.
- W3093671646 hasConceptScore W3093671646C76155785 @default.
- W3093671646 hasConceptScore W3093671646C82876162 @default.
- W3093671646 hasConceptScore W3093671646C93996380 @default.
- W3093671646 hasLocation W30936716461 @default.
- W3093671646 hasOpenAccess W3093671646 @default.
- W3093671646 hasPrimaryLocation W30936716461 @default.
- W3093671646 hasRelatedWork W10202958 @default.
- W3093671646 hasRelatedWork W10342353 @default.
- W3093671646 hasRelatedWork W12793662 @default.
- W3093671646 hasRelatedWork W13678974 @default.
- W3093671646 hasRelatedWork W13846533 @default.
- W3093671646 hasRelatedWork W1674447 @default.
- W3093671646 hasRelatedWork W6468916 @default.
- W3093671646 hasRelatedWork W7303821 @default.
- W3093671646 hasRelatedWork W8021486 @default.
- W3093671646 hasRelatedWork W9190101 @default.
- W3093671646 isParatext "false" @default.
- W3093671646 isRetracted "false" @default.
- W3093671646 magId "3093671646" @default.
- W3093671646 workType "article" @default.