This paper prepared by ETH Zürich was accepted in journal Future Generation of Computer Systems.
Highlights
- In-network offload can improve allreduce performance up to 2x.
- Existing approaches are prone to congestion due to static network paths selection.
- Canary is the first congestion-aware in-network allreduce algorithm.
- Canary improves allreduce performance by up to 40% compared to state-of-the-art.
Abstract
The allreduce operation is an essential building block for many distributed applications, ranging from the training of deep learning models to scientific computing. In an allreduce operation, data from multiple hosts is aggregated together and then broadcasted to each host participating in the operation. Allreduce performance can be improved by a factor of two by aggregating the data directly in the network. Switches aggregate data coming from multiple ports before forwarding the partially aggregated result to the next hop. In all existing solutions, each switch needs to know the ports from which it will receive the data to aggregate. However, this forces packets to traverse a predefined set of switches, making these solutions prone to congestion. For this reason, we design Canary, the first congestion-aware in-network allreduce algorithm. Canary uses load balancing algorithms to forward packets on the least congested paths. Because switches do not know from which ports they will receive the data to aggregate, they use timeouts to aggregate the data in a best-effort way. We develop a P4 Canary prototype and evaluate it on a Tofino switch. We then validate Canary through simulations on large networks, showing performance improvements up to 40% compared to the state-of-the-art.
Authors
Daniele De Sensi (ETH Zürich / Sapienza University of Rome), Edgar Costa Molero (ETH Zürich), Salvatore Di Girolamo (ETH Zürich), Laurent Vanbever (ETH Zürich), Torsten Hoefler (ETH Zürich)