Maximizing Utilization and Performance of Guaranteed-Bandwidth Long Fat Networks and Virtual Circuits
D. Michael Freemon

TL;DR
This paper demonstrates that using Hierarchical Token Bucket (HTB) without TCP congestion control over guaranteed-bandwidth virtual circuits significantly improves data transfer performance for large-scale scientific projects like LSST, especially over long fat networks.
Contribution
It introduces the use of HTB alone for bandwidth management over guaranteed virtual circuits, bypassing TCP congestion control to enhance throughput in long fat networks.
Findings
HTB effectively manages bandwidth for diverse traffic types.
Using HTB without TCP congestion control achieves near wire-speed throughput.
The approach addresses traditional TCP performance issues over long fat networks.
Abstract
Like many big science projects, the Large Synoptic Survey Telescope (LSST) has multiple geographic locations among which large amounts of data must be transferred. One particular type of data, crosstalk-corrected images, must be moved from South America to North America under stringent deadline requirements. LSST is provisioning an international network with bandwidth guarantees to handle this traffic. In prior work, we re-examined TCP congestion control for this use case and found that TCP throughput can approach wire speeds. This work shows that the Hierarchical Token Bucket (HTB) provides an excellent mechanism by which bandwidth can be managed for a wide range of traffic types. Using HTB without TCP congestion control over guaranteed-bandwidth virtual circuits is a compelling solution to the historical problem of poor TCP performance over long fat networks.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsInterconnection Networks and Systems · Parallel Computing and Optimization Techniques · Network Traffic and Congestion Control
