# Efficient Data-Plane Memory Scheduling for In-Network Aggregation

arXiv: https://arxiv.org/abs/2201.06398 · HTML: https://mlsanity.com/paper/2201.06398/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 52 references that are in the map.
- First posted on arXiv 17 Jan 2022.

## Builds on (most cited first, 10 of 10)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://mlsanity.com/paper/1409.1556/) (2014)
- [SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems](https://mlsanity.com/paper/1905.00537/) (2019)
- [Analysis of Large-Scale Multi-Tenant GPU Clusters for DNN Training Workloads](https://mlsanity.com/paper/1901.05758/) (2019)
- [Scaling Distributed Machine Learning with In-Network Aggregation](https://mlsanity.com/paper/1903.06701/) (2019)
- [DeepLight: Learning Illumination for Unconstrained Mobile Mixed Reality](https://mlsanity.com/paper/1904.01175/) (2019)
- [Poseidon: An Efficient Communication Architecture for Distributed Deep Learning on GPU Clusters](https://mlsanity.com/paper/1706.03292/) (2017)
- [Domain-specific Communication Optimization for Distributed DNN Training](https://mlsanity.com/paper/2008.08445/) (2020)
- [Programmable Switch as a Parallel Computing Device](https://mlsanity.com/paper/1803.01491/) (2018)
- [A Survey on Data Plane Programming with P4: Fundamentals, Advances, and Applied Research](https://mlsanity.com/paper/2101.10632/) (2021)

## Abstract (arXiv)

As the scale of distributed training grows, communication becomes a bottleneck. To accelerate the communication, recent works introduce In-Network Aggregation (INA), which moves the gradients summation into network middle-boxes, e.g., programmable switches to reduce the traffic volume. However, switch memory is scarce compared to the volume of gradients transmitted in distributed training. Although literature applies methods like pool-based streaming or dynamic sharing to tackle the mismatch, switch memory is still a potential performance bottleneck. Furthermore, we observe the under-utilization of switch memory due to the synchronization requirement for aggregator deallocation in recent works. To improve the switch memory utilization, we propose ESA, an underlineEfficient Switch Memory underlineScheduler for In-Network underlineAggregation. At its cores, ESA enforces the preemptive aggregator allocation primitive and introduces priority scheduling at the data-plane, which improves the switch memory utilization and average job completion time (JCT). Experiments show that ESA can improve the average JCT by up to 1.35×.