Shift Parallelism Reproduction Datasets
Abstract
These traces are derived from 15-minute segments from the original open-source production traces: Azure LLM Code TraceWe skip the first three minutes (which is mostly silent), and take the next 2834 requests of the original dataset. Mooncake Conversation TraceWe take the first 15 minutes, corresponding to the first 2728 requests. The derived datasets are given in in jsonl format and each row corresponds to a request that has three fields: "timestamp", "input_length", and "output_length".
Fields of Study & Subjects
Parallel Computing and Optimization Techniques
Cloud Computing and Resource Management
Software System Performance and Reliability
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 14 January 2026
- Document type
- Dataset