Rujia Wang

Redwan Ibne Seraj Khan, Kunal Jain, Haiying Shen et al.

In a multi-tenant large language model (LLM) serving platform hosting diverse applications, some users may submit an excessive number of requests, causing the service to become unavailable to other users and creating unfairness. Existing fairness approaches do not account for variations in token lengths across applications and multiple LLM calls, making them unsuitable for such platforms. To address the fairness challenge, this paper analyzes millions of requests from thousands of users on MS CoPilot, a real-world multi-tenant LLM platform hosted by Microsoft. Our analysis confirms the inadequacy of existing methods and guides the development of FairServe, a system that ensures fair LLM access across diverse applications. FairServe proposes application-characteristic aware request throttling coupled with a weighted service counter based scheduling technique to curb abusive behavior and ensure fairness. Our experimental results on real-world traces demonstrate FairServe's superior performance compared to the state-of-the-art method in ensuring fairness. We are actively working on deploying our system in production, expecting to benefit millions of customers world-wide.

Yizhu Jiao, Xuchao Zhang, Zhaoyang Wang et al.

Current Large Language Models (LLMs) excel in general reasoning yet struggle with specialized tasks requiring proprietary or domain-specific knowledge. Fine-tuning large models for every niche application is often infeasible due to black-box constraints and high computational overhead. To address this, we propose a collaborative framework that pairs a specialized weak model with a general strong model. The weak model, tailored to specific domains, produces initial drafts and background information, while the strong model leverages its advanced reasoning to refine these drafts, extending LLMs' capabilities to critical yet specialized tasks. To optimize this collaboration, we introduce a collaborative feedback to fine-tunes the weak model, which quantifies the influence of the weak model's contributions in the collaboration procedure and establishes preference pairs to guide preference tuning of the weak model. We validate our framework through experiments on three domains. We find that the collaboration significantly outperforms each model alone by leveraging complementary strengths. Moreover, aligning the weak model with the collaborative preference further enhances overall performance.

Zhou Ni, Rujia Wang

Secure multi-party computation (MPC) is a broad cryptographic concept that can be adopted for privacy-preserving computation. With MPC, a number of parties can collaboratively compute a function, without revealing the actual input or output of the plaintext to others. The applications of MPC range from privacy-preserving voting, arithmetic calculation, and large-scale data analysis. From the system perspective, each party in MPC can run on one compute node. The compute nodes of multiple parties could be either homogeneous or heterogeneous; however, the distributed workloads from the MPC protocols tend to be always homogeneous (symmetric). In this work, we study a representative MPC framework and a set of MPC applications from the system performance perspective. We show the detailed online computation workflow of a state-of-the-art MPC protocol and analyze the root cause of its stall time and performance bottleneck on homogeneous and heterogeneous compute nodes.