Elastic High Performance Computing (E-HPC) is a computing service provided by Alibaba Cloud that delivers excellent and stable computing performance through auto scaling and parallel computing. E-HPC aggregates computing capabilities to address major challenges in science, engineering, and commerce.
E-HPC is used across industries including scientific research, petroleum exploration, finance, weather forecasting, pharmaceuticals, gene sequencing, and graphics processing.
How it works
An E-HPC cluster runs on Elastic Compute Service (ECS) instances organized into three node types:
Logon node: Install, compile, and debug software, and submit jobs.
Management node: Hosts the scheduler server (which runs the scheduler client and manages scheduling nodes) and the domain account server (which manages cluster users).
Compute node: Runs HPC jobs. Overall cluster performance depends on compute node configuration.
The cluster runs inside a Virtual Private Cloud (VPC) for network isolation. The logon node is associated with an Elastic IP address (EIP) and is allocated a specific amount of public bandwidth to access the Internet. A security group—a virtual firewall with Stateful Packet Inspection (SPI) and packet filtering—controls inbound and outbound traffic to isolate security domains.
File Storage NAS (NAS) provides shared file access across all nodes. Object Storage Service (OSS) stores job data files and executable files, which you can import directly into the cluster.The file storage
Features
| Feature | Description | References |
|---|---|---|
| Computing resource creation | Create a cluster as the foundation for submitting and running jobs. E-HPC supports cluster creation and hybrid cloud cluster creation. | Create a cluster, Create a hybrid cloud cluster |
| Computing resource configuration | Configure users, nodes, scheduler queues, and auto scaling settings to match your workload requirements. | User management, Node management, Configure auto scaling |
| Job submission | Submit jobs from the E-HPC console or via SSH. For small jobs, edit and submit the job script directly in the console. For complex jobs, upload the script to an OSS bucket and import it into the console. | Job management, Create a job |
| Performance analysis | The E-HPC optimizer is a performance monitoring and profiling platform. Monitor cluster performance metrics, review performance history, and profile performance hotspots. | View cluster performance |
| Virtual Network Computing (VNC) | Use the VNC service to manage jobs from the E-HPC console. | Use VNC to manage a visualization service |