Get started with E-HPC Instant Compute Service

Prerequisites

You have activated E-HPC Instant Compute Service.

Workflow

Step 1: Add an application

E-HPC Instant Compute Service provides public applications, including mainstream scientific computing software, compiler runtime libraries, and MPI communication libraries. If the application or version you need is not in the public application list, you must add it.

1. Log on to the Instant Computing Service console - My Images and click Add image.

2. Configure the image settings and click OK.

   1. ImageType:

      • Virtual Machine Mirroring: A self-contained unit that includes a complete operating system and all programs.

      • Container Image: A minimal environment required to run a specific program.

   2. Image: If you select a virtual machine image, select an image from the drop-down list. If you select a Container Image, enter the container image address.

Step 2: Create a job

1. For the application you want to use, click Use.

   You are automatically redirected to the Create job page with the application pre-selected. Alternatively, go to the Jobs page of the E-HPC Instant console and click Create Job.

2. On the Create Job page, configure the following parameters.

   • Job information

     Parameter	Required	Description
     Job Name	Yes	Enter a name for the job. The name must meet the following requirements: Be 2 to 64 characters long. Start with a letter. Contain only letters, digits, and hyphens (-).
     Job Description	No	Enter a description of the job to help you identify and manage it. The description is displayed on the job details page.
     Area	Yes	Select the region where the job will run. This selection affects network latency and cost.
     Scheduling Policy	Yes	Select a scheduling policy for the job. Currently, only the Standard policy is supported. Note To use the Economy scheduling policy, submit a ticket.
     Proprietary Network	Conditionally required	In Network and Availability Zone, select the VPC to which the execution nodes belong. This parameter is required if you want to mount a NAS or CPFS file system, or if the execution nodes need to access resources in your VPC.
     Switch	Conditionally required	Select the vSwitch to which the execution nodes belong. You can filter vSwitches by zone. After you select a vSwitch, E-HPC Instant automatically attaches an elastic network interface (ENI) to each execution node so that the node can communicate with your data stores, such as NAS file systems.
     Security Group	No	After you select a VPC, the system automatically selects a security group in that VPC. You can select another security group. Multiple security groups are supported.
     Single-node Public EIP	No	A switch, which is turned off by default. After you turn it on, an elastic IP address (EIP) is automatically requested and assigned to each execution node so that the node can be accessed over the Internet. Important Only ECS instances are supported. Containers are not supported. Enabling an elastic IP address (EIP) incurs additional charges. We recommend that you enable Single-node Public EIP only when necessary.
     Tags	No	Select one or more tags to simplify resource searches and cost analysis. For more information, see Tags.

   • Task Details

     Parameter	Required	Description
     Computing Environment	Yes	Select the computing environment for the job. Options include: Cloud Server (default): runs job scripts on ECS instances. Container: runs the job in a container. You must use a container image created in Image Registry and configure additional parameters such as Image Pull Policy.
     Image	Yes	Select the image in which the job runs. The image serves as the computing environment of the job. Cloud Server: you can select a public image provided by the platform. The available public images are subject to what is displayed in the console. Container: you must select a container image that you created in Image Registry. If no image is available, click Create image to create one in Image Registry.
     Image Pull Policy	No	This parameter is displayed only when Computing Environment is set to Container. The following policies are supported: Pull If Not Present (default). Always Pull.
     Maximum Image Pull Timeout	No	This parameter is displayed only when Computing Environment is set to Container. Set the maximum amount of time to wait for the image to be pulled. Unit: minutes.
     Operation mode	Yes	Select the execution mode for the job. Two modes are available: Batch Job (default): runs the submitted commands and automatically releases the execution nodes after the job is complete. Background Service: runs the submitted commands and supports long-running tasks. The execution nodes are not released after the job is complete. You must manually delete the job to release the nodes. In this mode, Run Command is not displayed.
     Pre-processing script	No	Enter the script commands that you want to run before the job starts. This parameter is displayed only when Computing Environment is set to Cloud Server.
     Run Command	Yes	This parameter is displayed only when Operation mode is set to Batch Job. Enter the command or script to run on the job nodes. This parameter is required. For Linux applications, the command must start with #! followed by an interpreter, such as #!/bin/bash. For Windows applications, the first line of the command must be [powershell] or [bat]. Container: the startup command is split by spaces. You can use Parameter to append arguments to the command.
     Parameter	No	This parameter is displayed only when Computing Environment is set to Container. Click Create Parameter to append arguments to the container startup command, for example, --port=80.
     Number of execution nodes	Yes	The total number of nodes required to run the job. Valid values: 1 to 100. Default value: 1.

   • Task Resources

     The following two configuration methods are supported: Unified Computing (default; specifies computing power by combining the computing power level, vCPUs per node, and memory per node) and Select Type (directly selects ECS instance types).

     Parameter	Required	Description
     Computing Power	Yes	This parameter is required only for the Unified Computing method. Select the computing power level. Only General-purpose.
     Single-node cores	Yes	This parameter is required only for the Unified Computing method. Select the number of vCPUs for each execution node based on your business needs. Valid values: 2, 4, 6, 8, 10, 12, 14, and 16.
     Single node memory size	Yes	This parameter is required only for the Unified Computing method. Select the memory size of each execution node. The valid values are linked to the number of vCPUs per node and can be 1, 2, 4, 6, or 8 times the number of vCPUs. For example, if the number of vCPUs is 2, you can select 2 GiB, 4 GiB, 8 GiB, 12 GiB, or 16 GiB. Note The ratio of CPU cores to memory size for a single node is fixed. For more information, see Limits.
     Instance Type	Yes	This parameter is required only for the Select Type method. Select the ECS instance types for running the job. You can select multiple instance types. You can filter instance types by vCPUs, memory, architecture, and category, and sort the results. The list shows the instance family, vCPUs, memory, and reference price of each instance type.
     Storage Space per Node	Yes	Set the system disk size of each execution node. The default and minimum value is 20 GiB. The selected storage size must be larger than the image size. If you use a public image, we recommend that you set this parameter to 40 GiB or larger.
     Resource Pool	Yes	Select the resource pool in which the job runs. The default value is defaultpool. Jobs are scheduled based on the priority and execution node limit of the selected resource pool. You can create and manage resource pools on the Resource Scheduling Management > Resource Pool page in the left-side navigation pane.

   • Access External Storage

     Mount external storage for the job. While the job is running, you can read and write files in the mount directory as if they were local files. Click Add More to configure multiple storage mounts for the same job. To remove a mount that is no longer needed, click Delete.

     Parameter	Description
     Mount Directory	Enter the mount directory on the execution node. The default value is /mnt. Job scripts access the mounted storage through this directory.
     Type	Select the type of storage to mount. The supported types are General-purpose NAS, High-Performance CPFS, and OSS. Different storage types require different parameters. Note If no VPC or vSwitch is selected, General-purpose NAS and High-Performance CPFS cannot be selected. Complete the configuration in Network and Availability Zone first.
     File System	This parameter is displayed only when Type is set to General-purpose NAS or High-Performance CPFS. Select the NAS or CPFS file system to mount. If no file system is available, click Create File System to go to the console of the corresponding service and create one.
     Mount Point	This parameter is displayed only when Type is set to General-purpose NAS or High-Performance CPFS. Select a mount target of the file system. Mount targets are bound to VPCs, so only the mount targets in the selected VPC are displayed. Make sure that the file system has available mount target quota. If no mount target is available in the drop-down list, click Create Mount Target to create a mount target in the same VPC as the job for the file system.
     File System Directory	This parameter is displayed only when Type is set to General-purpose NAS or High-Performance CPFS. Enter the directory of the file system to mount. The directory must already exist in the file system.
     Mount Options	Select the mount protocol. NFS v3 (default) and NFS v4 are supported. This parameter is required only for General-purpose NAS.
     Bucket	Select the OSS bucket to mount. The default value is Bucket of Current Account. The drop-down list shows the buckets that belong to the current account in the current region. To mount a bucket of another account, switch to Custom Cross-account Bucket and enter the bucket name manually. This parameter is displayed only when Type is set to OSS.
     Subdirectory	Enter the subdirectory of the bucket to mount. The default value is the root directory (/). This parameter is displayed only when Type is set to OSS.
     AccessKey	Enter the AccessKey ID that is used to access the bucket. The AccessKey pair is used only to access the specified bucket directory. We recommend that you strictly control permissions related to the key pair.
     AccessKey Secret	Enter the AccessKey secret that corresponds to the preceding AccessKey ID. Important The Alibaba Cloud account or RAM user to which the AccessKey pair belongs must have read and write permissions on the bucket, and must not be denied oss:ListObjects. Before mounting, the system verifies the bucket by calling ListObjects. If the verification fails, the mount does not take effect and the console does not report an error. In this case, the mount directory becomes a local directory on the execution node, the data that you write is not saved to the bucket, and the data is lost when the execution node is released.

   • More Settings

     Parameter	Description
     Logon Password	If you need to log on to the nodes remotely, you must set the Logon Password and Confirm Password. Note You can set a password only when the Computing Environment is set to Cloud Server.
     Confirm Password
     Resource scheduling limit	This parameter is displayed only after you select a VPC. The following scheduling methods are supported: None (default): resources can be scheduled in the specified VPC. Specific VSwitch only: Creates execution nodes only in the specified vSwitch.
     Hyper-Threading	A switch, which is turned on by default. In some HPC scenarios, you must disable CPU hyper-threading to obtain better performance. Before you disable hyper-threading, make sure that the selected instance types support this operation. For more information, see Supported instance types.
     Environment Variables	Click Add Environment Variable to set environment variables for the job. You must specify a variable name and a variable value. Job scripts can directly reference these environment variables.
     Hostname Prefix	Set the hostname prefix of the execution nodes. The hostname of an execution node consists of the prefix and a sequence number in the format of prefix-number. For example, if the prefix is doctest, the hostname of the first execution node is doctest-0. If you do not set this parameter, the system randomly generates hostnames. The naming rules are as follows: Linux: The hostname can be up to 32 characters in length. Each segment can contain only letters, digits, and hyphens (-). A segment cannot start or end with a period (.) or a hyphen (-), and periods and hyphens cannot be used consecutively. Windows: The hostname can be up to 10 characters in length and can contain only letters, digits, and hyphens (-). It cannot consist of only digits, and periods (.) are not supported.
     Job Priority	Set the scheduling priority of the job. Valid values: integers from 1 to 100. A larger value indicates a higher scheduling priority. Default value: 1.
     Save as Job Template	A switch, which is turned off by default. After you turn it on, the current job configuration is saved as a job template. You can view and reuse templates on the Job Templates page in the left-side navigation pane.

3. After you configure the parameters, click Create Job at the bottom of the page.

Step 3: View job details

1. After the job is created, it appears in the job list. You can filter jobs by tag, status, or time.

2. Click the target job ID in the Job ID/Name column, or click View Details in the Operation column.

3. On the job details page, you can view the job’s basic information, task execution status, and the status of each execution node.

   • Basic information: Displays details such as the job ID, status, region, network, security group, image ID, resource allocation policy, computing environment, mode, creation/start/end time, and computing power level.

   • Execution node status: Displays information about the job's execution nodes, including the Total number of tasks performed, Number of running tasks, Number of tasks completed, and Terminated Tasks.

   • Task execution status: Displays a list of tasks with columns for Type, Cores, Memory, and Compute Storage Space. You can expand an instance to view its details, including its sequence number, status, start/end time, and hostname. The Operation column provides links to Remote Login, Monitoring, and Log.

4. View the following job execution details:

   • Monitoring data: Includes CPU utilization, memory utilization, and disk usage.

   • Runtime logs: Includes the runtime logs for the pre-execution script and the execution command.

   • Remote login: If an issue occurs, you can log in to an execution node to troubleshoot.