4 papers
Trivance: Latency-Optimal AllReduce by Shortcutting Multiport Networks
Anton Juerss, Vamsi Addanki, Stefan Schmid
AllReduce is a fundamental collective communication operation in distributed computing and a key performance bottleneck for large-scale training and inference. Its completion time…
The Harmonic Policy for Online Buffer Sharing is (2 + ln n)-Competitive: A Simple Proof
Vamsi Addanki, Julien Dallot, Leon Kellerhals +2
The problem of online buffer sharing is expressed as follows. A switch with output ports receives a stream of incoming packets. When an incoming packet is accepted by the switc…
Short-circuiting Rings for Low-Latency AllReduce
Sarah-Michelle Hammer, Stefan Schmid, Rachee Singh +1
Efficient collective communication is critical for many distributed ML and HPC applications. In this context, it is widely believed that the Ring algorithm for the AllReduce collec…
Learning Minimum Linear Arrangement of Cliques and Lines
Julien Dallot, Maciej Pacut, Marcin Bienkowski +2
In the well-known Minimum Linear Arrangement problem (MinLA), the goal is to arrange the nodes of an undirected graph into a permutation so that the total stretch of the edges is m…