Lingjun uses clusters, node groups, and compute nodes to organize high-performance computing resources. You can plan multiple clusters based on your business requirements and divide them into node groups for efficient resource utilization.
|
Concept |
Description |
|
cluster |
A collection of high-performance heterogeneous accelerated compute nodes equipped with Lingjun optimization components. Nodes in a cluster communicate over an 800 Gbps high-speed, low-latency RDMA network. You can use native physical cluster services or combine them with other Alibaba Cloud products. |
|
Nodes |
High-performance GPU servers accelerated by Lingjun optimization components. You can select the operating system (OS) deployed on compute nodes. Currently, CentOS 7.9 is supported. |
|
Group |
A collection of nodes that forms a subset of a cluster. A group typically consists of one or more nodes with the same specifications and features, such as a GU100 node group. |
|
Optimization Suite |
An optimization suite for large-scale parallel compute clusters. The suite includes data loading optimization, collective communication optimization, compute resource optimization, and network optimization. |