Score: 0

SpecEdge: Scalable Edge-Assisted Serving Framework for Interactive LLMs

Published: May 16, 2025 | arXiv ID: 2505.17052v1

By: Jinwoo Park, Seunggeun Cho, Dongsu Han

Potential Business Impact:

Makes AI run faster and cheaper everywhere.

Business Areas:
Semantic Web Internet Services

Large language models (LLMs) power many modern applications, but serving them at scale remains costly and resource-intensive. Current server-centric systems overlook consumer-grade GPUs at the edge. We introduce SpecEdge, an edge-assisted inference framework that splits LLM workloads between edge and server GPUs using a speculative decoding scheme, exchanging only token outputs over the network. SpecEdge employs proactive edge drafting to overlap edge token creation with server verification and pipeline-aware scheduling that interleaves multiple user requests to increase server-side throughput. Experiments show SpecEdge enhances overall cost efficiency by 1.91x through achieving 2.22x server throughput, and reduces inter token latency by 11.24% compared to a server-only baseline, introducing a scalable, cost-effective paradigm for LLM serving.

Country of Origin
🇰🇷 Korea, Republic of

Page Count
17 pages

Category
Computer Science:
Computation and Language