Alibaba Cloud’s Mint Tracing Framework and FAMOS Diagnosis Earn Top‑Conference Spot

Mint: Cost‑Efficient Tracing Framework (ASPLOS)

In the ASPLOS‑selected paper “Mint: Cost‑Efficient Tracing with All Requests Collection via Commonality and Variability Analysis,” Alibaba Cloud introduces Mint, a novel tracing framework that adopts a “commonality + variability” sampling paradigm. By separating shared behaviors from request‑specific differences, Mint retains essential information while dramatically reducing data volume, achieving average storage reduction to 2.7% and network overhead reduction to 4.2% without sacrificing trace completeness.

Traditional tracing systems use binary sampling—either record everything or almost nothing—leading to either overwhelming data or insufficient insight. Mint’s approach intelligently aggregates common parts of request traces and filters out redundant variations, akin to preserving the opening and closing of many play scripts while distinguishing their unique middle scenes.

FAMOS: Fault Diagnosis for Microservice Systems (ICSE)

The ICSE‑selected paper “FAMOS: Fault diagnosis for Microservice Systems through Effective Multi‑modal Data Fusion” presents a fault‑diagnosis method named FAMOS. It tackles the difficulty of locating issues in complex IT environments where single data sources (logs, metrics, traces) are insufficient. FAMOS designs optimal feature extraction for each data type, fuses them across modalities, and captures inter‑source correlations, greatly improving diagnosis accuracy and efficiency.

The method is illustrated like a detective combining witness testimony, photos, and video footage to reconstruct a case, enabling a more complete understanding of system failures.

Impact and Integration

Both research outcomes have already been partially integrated into Alibaba Cloud’s observability product family, such as the Log Service (SLS) and Application Real‑Time Monitoring Service (ARMS). Real‑world customers—e.g., a gaming company monitoring global launches, a beverage chain improving incident recovery by over 50%, and an automotive brand shortening alert resolution time by 50%—benefit from these advances.

ASPLOS and ICSE are top‑tier international conferences in software engineering and computer systems, with acceptance rates of roughly 12.7% and 15‑20% respectively, underscoring the significance of Alibaba Cloud’s contributions to the observability field.