Just in time delivery: Leveraging operating systems knowledge for better datacenter congestion control
Network links and server CPUs are heavily contended resources in modern datacenters. To keep tail latencies low, datacenter operators drastically overprovision both types of resources today, and there has been significant research into effectively managing network traffic [4, 19, 21, 29] and CPU loa...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
USENIX Association
2020
|
| Online Access: | https://hdl.handle.net/1721.1/128726 |
| _version_ | 1826198440173895680 |
|---|---|
| author | Ousterhout, Amy Elizabeth Belay, Adam M Zhang, I |
| author2 | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory |
| author_facet | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Ousterhout, Amy Elizabeth Belay, Adam M Zhang, I |
| author_sort | Ousterhout, Amy Elizabeth |
| collection | MIT |
| description | Network links and server CPUs are heavily contended resources in modern datacenters. To keep tail latencies low, datacenter operators drastically overprovision both types of resources today, and there has been significant research into effectively managing network traffic [4, 19, 21, 29] and CPU load [22, 27, 32]. However, this work typically looks at the two resources in isolation. In this paper, we make the observation that, in the datacenter, the allocation of network and CPU resources should be co-designed for the most efficiency and the best response times. For example, while congestion control protocols can prioritize traffic from certain flows, this provides no benefit if the traffic arrives at an overloaded server that will only queue the request. This paper explores the potential benefits of such a co-designed resource allocator and considers the recent work in both CPU scheduling and congestion control that is best suited to such a system. We propose a Chimera, a new datacenter OS that integrates a receiver-based congestion control protocol with OS insight into application queues, using the recent Shenango operating system [32]. |
| first_indexed | 2024-09-23T11:04:59Z |
| format | Article |
| id | mit-1721.1/128726 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T11:04:59Z |
| publishDate | 2020 |
| publisher | USENIX Association |
| record_format | dspace |
| spelling | mit-1721.1/1287262022-09-27T17:00:50Z Just in time delivery: Leveraging operating systems knowledge for better datacenter congestion control Ousterhout, Amy Elizabeth Belay, Adam M Zhang, I Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Network links and server CPUs are heavily contended resources in modern datacenters. To keep tail latencies low, datacenter operators drastically overprovision both types of resources today, and there has been significant research into effectively managing network traffic [4, 19, 21, 29] and CPU load [22, 27, 32]. However, this work typically looks at the two resources in isolation. In this paper, we make the observation that, in the datacenter, the allocation of network and CPU resources should be co-designed for the most efficiency and the best response times. For example, while congestion control protocols can prioritize traffic from certain flows, this provides no benefit if the traffic arrives at an overloaded server that will only queue the request. This paper explores the potential benefits of such a co-designed resource allocator and considers the recent work in both CPU scheduling and congestion control that is best suited to such a system. We propose a Chimera, a new datacenter OS that integrates a receiver-based congestion control protocol with OS insight into application queues, using the recent Shenango operating system [32]. 2020-12-03T18:46:07Z 2020-12-03T18:46:07Z 2019-07 2020-12-01T18:33:54Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/128726 Ousterhout, Amy et al. "Just in time delivery: Leveraging operating systems knowledge for better datacenter congestion control." 11th USENIX Workshop on Hot Topics in Cloud Computing, HotCloud 2019, co-located with USENIX ATC 2019, July 2019, Renton, Washington, USENIX Association, July 2019. © 2019 USENIX Association en https://www.usenix.org/conference/hotcloud19/presentation/ousterhout 11th USENIX Workshop on Hot Topics in Cloud Computing, HotCloud 2019, co-located with USENIX ATC 2019 Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf USENIX Association other univ website |
| spellingShingle | Ousterhout, Amy Elizabeth Belay, Adam M Zhang, I Just in time delivery: Leveraging operating systems knowledge for better datacenter congestion control |
| title | Just in time delivery: Leveraging operating systems knowledge for better datacenter congestion control |
| title_full | Just in time delivery: Leveraging operating systems knowledge for better datacenter congestion control |
| title_fullStr | Just in time delivery: Leveraging operating systems knowledge for better datacenter congestion control |
| title_full_unstemmed | Just in time delivery: Leveraging operating systems knowledge for better datacenter congestion control |
| title_short | Just in time delivery: Leveraging operating systems knowledge for better datacenter congestion control |
| title_sort | just in time delivery leveraging operating systems knowledge for better datacenter congestion control |
| url | https://hdl.handle.net/1721.1/128726 |
| work_keys_str_mv | AT ousterhoutamyelizabeth justintimedeliveryleveragingoperatingsystemsknowledgeforbetterdatacentercongestioncontrol AT belayadamm justintimedeliveryleveragingoperatingsystemsknowledgeforbetterdatacentercongestioncontrol AT zhangi justintimedeliveryleveragingoperatingsystemsknowledgeforbetterdatacentercongestioncontrol |