Publications
Detailed Information
Layerweaver plus : A QoS-Aware Layer-Wise DNN Scheduler for Multi-Tenant Neural Processing Units
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Oh, Young H. | - |
dc.contributor.author | Jin, Yunho | - |
dc.contributor.author | Ham, Tae Jun | - |
dc.contributor.author | Lee, Jae W. | - |
dc.date.accessioned | 2022-05-04T01:43:15Z | - |
dc.date.available | 2022-05-04T01:43:15Z | - |
dc.date.created | 2022-02-14 | - |
dc.date.issued | 2022-01-01 | - |
dc.identifier.citation | IEICE Transactions on Information and Systems, Vol.E105D No.2, pp.427-431 | - |
dc.identifier.issn | 0916-8532 | - |
dc.identifier.uri | https://hdl.handle.net/10371/179351 | - |
dc.description.abstract | Many cloud service providers employ specialized hardware accelerators, called neural processing units (NPUs), to accelerate deep neural networks (DNNs). An NPU scheduler is responsible for scheduling incoming user requests and required to satisfy the two, often conflicting, optimization goals: maximizing system throughput and satisfying quality-of-service (QoS) constraints (e.g., deadlines) of individual requests. We propose Layerweaver+, a low-cost layer-wise DNN scheduler for NPUs, which provides both high system throughput and minimal QoS violations. For a serving scenario based on the industry-standard MLPerf inference benchmark, Layerweaver+ significantly improves the system throughput by up to 266.7% over the baseline scheduler serving one DNN at a time. | - |
dc.language | 영어 | - |
dc.publisher | Oxford University Press | - |
dc.title | Layerweaver plus : A QoS-Aware Layer-Wise DNN Scheduler for Multi-Tenant Neural Processing Units | - |
dc.type | Article | - |
dc.identifier.doi | 10.1587/transinf.2021EDL8084 | - |
dc.citation.journaltitle | IEICE Transactions on Information and Systems | - |
dc.identifier.wosid | 000748957000025 | - |
dc.identifier.scopusid | 2-s2.0-85124651814 | - |
dc.citation.endpage | 431 | - |
dc.citation.number | 2 | - |
dc.citation.startpage | 427 | - |
dc.citation.volume | E105D | - |
dc.description.isOpenAccess | N | - |
dc.contributor.affiliatedAuthor | Lee, Jae W. | - |
dc.type.docType | Article | - |
dc.description.journalClass | 1 | - |
- Appears in Collections:
- Files in This Item:
- There are no files associated with this item.
Item View & Download Count
Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.