CloudMirror: Application-Aware Bandwidth Reservations in the Cloud

on YouTube ↗27m

" summary

Proposes a graph-based tenant application graph that specifies per-VM directional send and receive guarantees and a new placement algorithm that maps that graph onto a tree datacenter to minimize core reservations. Simulations using the Microsoft bing.com communication traces on a 2,000-host, 50-VM-per-host model and comparisons to virtual cluster and VOC show up to 76 percent less total reserved bandwidth and often ~2x savings. Enforcement assumes existing tools like ElasticSwitch and vShaper and a planned OpenStack prototype.

AI-written summary · about

" chapters (19)

1. 0:00Intro
2. 0:32Need bandwidth guarantees for predictable performance
3. 2:10Goal: Network Abstraction for Expressing Bandwidth Demands
4. 3:09Cloud Mirror Context
5. 3:55Solution: Cloud Mirror
6. 4:49Pipe Model
7. 6:12Virtual Cluster Model
8. 6:59Virtual Cluster Example
9. 9:36Intuition: Model The Application Not The Network
10. 10:48Tenant Application Graph (TAG)
11. 11:48Bandwidth Models in TAG
12. 13:01TAG is Intuitive
13. 13:36TAG is Efficient
14. 15:15Flat Per-VM bandwidth while tenant scaling
15. 16:28VM Placement
16. 18:45Evaluations
17. 21:07Conclusion
18. 22:48Backup slides
19. 25:04Bandwidth demand is proportional to user transaction

" related talks

• 21mChatty Tenants and the Cloud Network Sharing ProblemChanghoon Kim, Hitesh Ballani, Keon Jang, Thomas Karagiannis, Dinan Gunawardena, Greg O'Shea · NSDI 2013
• 36mNetwork Bandwidth-Aware Kubernetes ClusterYifeng Xiao, Yang Yu · KubeCon + CloudNativeCon + Open Source Summit China 2019
• 10mLightning Talk: Closing the Gap: Fair North–South Bandwidth Management for TenantsWeizhou Lan · CiliumCon Europe 2026
• 17mNetHint: White-Box Networking for Multi-Tenant Data CentersWei Zhang, Ion Stoica, Jingrong Chen, Danyang Zhuo, Hong Zhang, Liang Luo, Jeffrey Chase · 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI '22)
• 16mPreview: Cloud ComputingKshiteej Mahajan · OSDI '21 / USENIX ATC '21
• 18mCloudcast: High-Throughput, Cost-Aware Overlay Multicast in the CloudIon Stoica, Paras Jain, Vincent Liu, Joseph E. Gonzalez, Sarah Wooders, Shu Liu, Xiangxi Mo · 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24)
• 16mOptimizing Network Provisioning through CooperationRanjita Bhagwan, Yogesh Bansal, Harsha Sharma, Parth Thakkar, Sagar Bharadwaj, Venkata N. Padmanabhan, Vijay Kumar, Kathleen Voelbel · 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI '22)
• 12mCost-effective Cloud Edge Traffic Engineering with CascaraSharad Agarwal, Paramvir Bahl, Rachee Singh, Matt Calder · NSDI ’21