The cloud uptime archive is a collection of failure traces from different cloud services. The cleaned and extracted traces are located in the traces folder. For some services, the traces have been collected from both the service operator and user reports to failure reporting websites such as Outage Report.
The traces folder contains a different csv file for each service-source pair. Each trace is in the following format:
| Start time | End time | Severity |
|---|---|---|
| timestamp | timestamp | value between 0-1 |
| ... | ... | ... |
Start time corresponds to the start time of the failure, and end time corresponds to the end time. Severity is a synthetic metric we generate based on the number of services and users affected. It can be interpreted as the probability that a task or a user request will fail in the time period represented by the start time and end time.
The scripts to analyze the data to produce the figures in the paper are in the analysis_scripts folder.
The CUA uses the OpenDC simulator for experiments. The experiment configurations we use are in the simulation_config folder.
The code to clean and extract failure traces from the data is also include in the repository, in the process_raw_data folder. However, the raw data itself is not included as it is large. The raw dataset is available on Zenodo.