Embedding Samples Dispatching for Recommendation Model Training in Edge Environments

2025-12-25

AI Summary

In edge environments, DLRM training suffers from large embedding tables and communication overhead between edge nodes and parameter servers, which dominates training latency. Method: This paper proposes the Embedding Sample Distribution (ESD) mechanism and HybridDis hybrid scheduling method. It is the first to jointly model network heterogeneity, edge resource constraints, and embedding sample transmission costs, enabling Pareto-optimal trade-offs between training quality and communication cost. The approach integrates optimal planning with lightweight heuristics to support dynamic sample distribution and collaborative caching. Results: Evaluated on real-world workloads, HybridDis reduces embedding transmission cost by up to 36.76% and achieves 1.74× end-to-end training speedup, significantly improving the efficiency of edge-based recommendation model training.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Distributed Search
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Training deep learning recommendation models (DLRMs) on edge workers brings several benefits, particularly in terms of data privacy protection, low latency and personalization. However, due to the huge size of embedding tables, typical DLRM training frameworks adopt one or more parameter servers to maintain global embedding tables, while leveraging the edge workers cache part of them. This incurs significant transmission cost for embedding transmissions between workers and parameter servers, which can dominate the training cycle. In this paper, we investigate how to dispatch input embedding samples to appropriate edge workers to minimize the total embedding transmission cost when facing edge-specific challenges such as heterogeneous networks and limited resources. We develop ESD, a novel mechanism that optimizes the dispatch of input embedding samples to edge workers based on expected embedding transmission cost. We propose HybridDis as the dispatch decision method within ESD, which combines a resource-intensive optimal algorithm and a heuristic algorithm to balance decision quality and resource consumption. We implement a prototype of ESD and compare it with state-of-the-art mechanisms on real-world workloads. Extensive experimental results show that ESD reduces the embedding transmission cost by up to 36.76% and achieves up to 1.74 times speedup in end-to-end DLRM training.

Problem

• Minimizes embedding transmission cost in edge DLRM training
• Addresses heterogeneous networks and limited edge resources
• Optimizes sample dispatch to reduce training cycle time

Innovation

• ESD optimizes embedding sample dispatch to edge workers
• HybridDis combines optimal and heuristic algorithms for dispatch
• ESD reduces embedding transmission cost by up to 36.76%