Anton A. Zabreyko
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Anton A. Zabreyko has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Manya Ghobadi
Massachusetts Institute of Technology
2 shared publications - Sanjoli Narang
Department of Electrical Engineering, IIT Bombay, Powai, Mumbai, India
1 shared publication - Sudarsanan Rajasekaran
Massachusetts Institute of Technology
1 shared publication - Weiyang Wang
MIT, Cambridge, MA, USA
1 shared publication
Latest Publications
- Open Access
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 1046-1061
MonkeyTree: Near-Minimal Congestion for Multi-tenant Training via Migration
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 1046-1061
Open AccessMonkeyTree: Near-Minimal Congestion for Multi-tenant Training via Migration
- Open Access
- 18 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 167-176
MLTCP: A Distributed Technique to Approximate Centralized Flow Scheduling For Machine Learning
This paper argues that congestion control protocols in machine learning datacenters sit at a sweet spot between centralized and distributed flow scheduling solutions. We present MLTCP, a technique to augment today's congestion control algorithms to approximate an interleaved centralized flow schedule. At the heart of MLTCP lies a straight-forward principle based on a key conceptual insight: by scaling the congestion window size (or sending rate) based on the number of bytes sent at each iteration, MLTCP flows eventually converge into a schedule that reduces network contention. We demonstrate that MLTCP uses a gradient descent trend with a step taken at every training (or fine-tuning) iteration towards reducing network congestion among competing jobs.
- 18 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 167-176
Open AccessMLTCP: A Distributed Technique to Approximate Centralized Flow Scheduling For Machine Learning
This paper argues that congestion control protocols in machine learning datacenters sit at a sweet spot between centralized and distributed flow scheduling solutions. We present MLTCP, a technique to augment today's congestion control algorithms to approximate an interleaved centralized flow schedule. At the heart of MLTCP lies a straight-forward principle based on a key conceptual insight: by scaling the congestion window size (or sending rate) based on the number of bytes sent at each iteration, MLTCP flows eventually converge into a schedule that reduces network contention. We demonstrate that MLTCP uses a gradient descent trend with a step taken at every training (or fine-tuning) iteration towards reducing network congestion among competing jobs.