SpliDT: Partitioned Decision Trees for Scalable Stateful Inference at Line Rate

2025-08-30

AI Summary

Existing data-plane decision trees are constrained by hardware resources, requiring precomputed, fixed-size feature sets—severely limiting model accuracy and scalability. This paper proposes a partitioned decision tree system for programmable data planes, enabling stateful streaming inference over sliding windows. Our approach addresses these limitations through three key contributions: (1) a subtree-based feature partitioning mechanism for distributed inference; (2) a loopback channel to enable cross-partition reuse of registers and match keys; and (3) a joint optimization framework that co-designs feature allocation and tree topology. Implemented in P4, the system integrates match-action tables, stateful registers, loopback control, and sliding-window feature extraction, supported by custom training and design-space exploration tools. Evaluation on real-world datasets shows it supports stateful feature sets five times larger than prior methods, achieves comparable detection latency, and incurs less than 0.05% loopback overhead under one million concurrent flows.

Technology Category

• Machine Learning: Hardware-aware ML
• Search and Optimization: Distributed Search
• Planning, Routing, and Scheduling: Planning/Scheduling and Learning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Machine learning (ML) is increasingly being deployed in programmable data planes (switches and SmartNICs) to enable real-time traffic analysis, security monitoring, and in-network decision-making. Decision trees (DTs) are particularly well-suited for these tasks due to their interpretability and compatibility with data-plane architectures, i.e., match-action tables (MATs). However, existing in-network DT implementations are constrained by the need to compute all input features upfront, forcing models to rely on a small, fixed set of features per flow. This significantly limits model accuracy and scalability under stringent hardware resource constraints. We present SPLIDT, a system that rethinks DT deployment in the data plane by enabling partitioned inference over sliding windows of packets. SPLIDT introduces two key innovations: (1) it assigns distinct, variable feature sets to individual sub-trees of a DT, grouped into partitions, and (2) it leverages an in-band control channel (via recirculation) to reuse data-plane resources (both stateful registers and match keys) across partitions at line rate. These insights allow SPLIDT to scale the number of stateful features a model can use without exceeding hardware limits. To support this architecture, SPLIDT incorporates a custom training and design-space exploration (DSE) framework that jointly optimizes feature allocation, tree partitioning, and DT model depth. Evaluation across multiple real-world datasets shows that SPLIDT achieves higher accuracy while supporting up to 5x more stateful features than prior approaches (e.g., NetBeacon and Leo). It maintains the same low time-to-detection (TTD) as these systems, while scaling to millions of flows with minimal recirculation overhead (<0.05%).

Problem

• Enables scalable stateful inference in data planes
• Overcomes hardware constraints on feature usage
• Supports partitioned decision trees for line rate

Innovation

• Partitioned decision trees for scalable stateful inference
• Variable feature sets assigned to individual sub-trees
• In-band control channel reuses data-plane resources