Alibaba Cluster Trace

Alibaba Cluster Trace captures detailed statistics for the co-located workloads of long-running and batch jobs over a course of 24 hours. The trace consists of three parts: (1) statistics of the studied homogeneous cluster of 1,313 machines, including each machine’s hardware configuration, and the runtime {CPU, Memory, Disk} resource usage for a duration of 12 hours (the 2nd half of the 24-hour period); (2) long-running job workloads, including a trace of all container deployment requests and actions, and a resource usage trace for 12 hours; (3) co-located batch job workloads, including a trace of all batch job requests and actions, and a trace of per-instance resource usage over 24 hours.

It also has a second version of traces cluster-trace-v2018 that includes about 4,000 machines in a period of 8 days. Besides having larger scaler than trace-v2017, this piece trace also contains the DAG information of the production batch workloads.