Introduction
BatchCompute is a distributed cloud service designed for large-scale, parallel batch processing jobs. It provides a comprehensive set of APIs to help you use the service and build your own applications.
Before using these APIs, ensure you understand the BatchCompute product documentation, terms of service, and billing methods.
BatchCompute limits the number and specifications of resources, such as clusters and jobs, available to each user. For more information, see Limits.
In case of any discrepancy between the parameter options or specifications in this API reference and the resource limits on the official website, the information on the official website prevails.
Usage notes
This API reference describes the request syntax, parameter definitions, and request and response examples. For rapid development, we recommend using the SDK. For installation and usage instructions, see the SDK Reference.
Glossary
Term | English | Description |
cluster | cluster | In BatchCompute, a cluster is a collection of instance groups. Instances within the same group share an instance type, such as a specific CPU and memory configuration, and all groups within the cluster share a common image. |
job | job | In BatchCompute, a job represents a specific computational workload. |
task | task | A job consists of a set of tasks and their dependencies. BatchCompute supports jobs structured as a directed acyclic graph (DAG). |
instance | instance | An instance is the smallest unit of work that BatchCompute schedules and executes. These instances run on virtual server nodes that the system dynamically allocates. A task can have one or more instances. The instances of the same task process their input data in parallel. |
image | image | An image is a template used to create cluster resources or define a job's runtime environment. It can be a standard or custom ECS image. The BatchCompute Agent must be installed on the image, and the BatchCompute service must be authorized to use it. |