All Products
Search
Document Center

Features

Last Updated: Nov 02, 2018

Submit jobs

You use a tool, such as an SDK or command line tool, to submit jobs to Batch Compute. Then, Batch Compute uses the specified image (such as Ubuntu) to start virtual machines (VMs). These VMs run your programs and are released after the job is completed.

  • In Batch Compute, OSS can be used for persistent storage. After a program is run, you can save the results data in OSS. In Batch Compute, you can also use file interfaces to access data in OSS. For more information, see OSS mounting function.

  • By default, Batch Compute programs run on VMs, but can also support Docker containers. This means that you can use custom ECS images or Docker to install any software required on your image and then use this image to run any program.

Job description

You must submit a job description file in JSON format to Batch Compute. This JSON files gives a detailed description of the programs that are to be run.This file also describes the number of hosts required to run the programs, the specifications of memory and CPU, where the operation log is to be printed, and the results output location.

A single job contains multiple tasks that are performed in the sequence described in the specified DAG.

dag-tasks

Each task defines its image, instance type, the program to run, the number of hosts, and the results storage location.

specifications

Manage jobs

You can use a tool (the console or command line tool) to view the jobs you have submitted and stop, restart, or delete jobs. You can also view the conditions of each task and the conditions and logs for each instance (VM). The console’s job management interface is shown in the following figure:

console-jobs

Use cluster

VMs must be started before a program is run, which takes a certain amount of time (generally several minutes). In addition, during busy periods, the resources you request may not be available (if the VMs are in use by other customers). Therefore, after submitting a job, you may have to wait some time before it can be run.

If you want to improve your operational efficiency, you can create a cluster before running jobs. To do this, specify the number of VMs you require, for example, five VMs, and your image ID. Then, Batch Compute allocates you the VMs and starts them. These VMs remain in the Running status. Once you have a cluster, your jobs are run as soon as you submit them, improving your efficiency.

Custom images

When you submit a job or create a cluster, you can use an official image provided by Batch Compute or your own custom image.Custom images allow you to install the software you need independently. For more information, click here.