Ce contenu n'est pas disponible dans la langue sélectionnée.
Chapter 5. NVIDIA GPU architecture overview
NVIDIA supports the use of graphics processing unit (GPU) resources on Red Hat OpenShift Service on AWS. Red Hat OpenShift Service on AWS is a security-focused and hardened Kubernetes platform developed and supported by Red Hat for deploying and managing Kubernetes clusters at scale. Red Hat OpenShift Service on AWS includes enhancements to Kubernetes so that users can easily configure and use NVIDIA GPU resources to accelerate workloads.
The NVIDIA GPU Operator leverages the Operator framework within Red Hat OpenShift Service on AWS to manage the full lifecycle of NVIDIA software components required to run GPU-accelerated workloads.
These components include the NVIDIA drivers (to enable CUDA), the Kubernetes device plugin for GPUs, the NVIDIA Container Toolkit, automatic node tagging using GPU feature discovery (GFD), DCGM-based monitoring, and others.
The NVIDIA GPU Operator is only supported by NVIDIA. For more information about obtaining support from NVIDIA, see Obtaining Support from NVIDIA.
5.1. NVIDIA GPU prerequisites
- A working OpenShift cluster with at least one GPU worker node.
-
Access to the OpenShift cluster as a
cluster-adminto perform the required steps. -
OpenShift CLI (
oc) is installed. -
The node feature discovery (NFD) Operator is installed and a
nodefeaturediscoveryinstance is created.
5.2. GPUs and ROSA
You can deploy Red Hat OpenShift Service on AWS on NVIDIA GPU instance types.
It is important that this compute instance is a GPU-accelerated compute instance and that the GPU type matches the list of supported GPUs from NVIDIA AI Enterprise. For example, T4, V100, and A100 are part of this list.
You can choose one of the following methods to access the containerized GPUs:
- GPU passthrough to access and use GPU hardware within a virtual machine (VM).
- GPU (vGPU) time slicing when the entire GPU is not required.
Additional resources
5.3. GPU sharing methods
Red Hat and NVIDIA have developed GPU concurrency and sharing mechanisms to simplify GPU-accelerated computing on an enterprise-level Red Hat OpenShift Service on AWS cluster.
Applications typically have different compute requirements that can leave GPUs underutilized. Providing the right amount of compute resources for each workload is critical to reduce deployment cost and maximize GPU utilization.
Concurrency mechanisms for improving GPU utilization exist that range from programming model APIs to system software and hardware partitioning, including virtualization. The following list shows the GPU concurrency mechanisms:
- Compute Unified Device Architecture (CUDA) streams
- Time-slicing
- CUDA Multi-Process Service (MPS)
- Multi-instance GPU (MIG)
- Virtualization with vGPU
Additional resources
5.3.1. CUDA streams
Compute Unified Device Architecture (CUDA) is a parallel computing platform and programming model developed by NVIDIA for general computing on GPUs.
A stream is a sequence of operations that executes in issue-order on the GPU. CUDA commands are typically executed sequentially in a default stream and a task does not start until a preceding task has completed.
Asynchronous processing of operations across different streams allows for parallel execution of tasks. A task issued in one stream runs before, during, or after another task is issued into another stream. This allows the GPU to run multiple tasks simultaneously in no prescribed order, leading to improved performance.
Additional resources
5.3.2. Time-slicing
GPU time-slicing interleaves workloads scheduled on overloaded GPUs when you are running multiple CUDA applications.
You can enable time-slicing of GPUs on Kubernetes by defining a set of replicas for a GPU, each of which can be independently distributed to a pod to run workloads on. Unlike multi-instance GPU (MIG), there is no memory or fault isolation between replicas, but for some workloads this is better than not sharing at all. Internally, GPU time-slicing is used to multiplex workloads from replicas of the same underlying GPU.
You can apply a cluster-wide default configuration for time-slicing. You can also apply node-specific configurations. For example, you can apply a time-slicing configuration only to nodes with Tesla T4 GPUs and not modify nodes with other GPU models.
You can combine these two approaches by applying a cluster-wide default configuration and then labeling nodes to give those nodes a node-specific configuration.
5.3.3. CUDA Multi-Process Service
CUDA Multi-Process Service (MPS) allows a single GPU to use multiple CUDA processes. The processes run in parallel on the GPU, eliminating saturation of the GPU compute resources. MPS also enables concurrent execution, or overlapping, of kernel operations and memory copying from different processes to enhance utilization.
Additional resources
5.3.4. Multi-instance GPU
Using Multi-instance GPU (MIG), you can split GPU compute units and memory into multiple MIG instances. Each of these instances represents a standalone GPU device from a system perspective and can be connected to any application, container, or virtual machine running on the node. The software that uses the GPU treats each of these MIG instances as an individual GPU.
MIG is useful when you have an application that does not require the full power of an entire GPU. The MIG feature of the new NVIDIA Ampere architecture enables you to split your hardware resources into multiple GPU instances, each of which is available to the operating system as an independent CUDA-enabled GPU.
NVIDIA GPU Operator version 1.7.0 and higher provides MIG support for the A100 and A30 Ampere cards. These GPU instances are designed to support up to seven multiple independent CUDA applications so that they operate completely isolated with dedicated hardware resources.
Additional resources
5.3.5. Virtualization with vGPU
Virtual machines (VMs) can directly access a single physical GPU using NVIDIA vGPU. You can create virtual GPUs that can be shared by VMs across the enterprise and accessed by other devices.
This capability combines the power of GPU performance with the management and security benefits provided by vGPU. Additional benefits provided by vGPU includes proactive management and monitoring for your VM environment, workload balancing for mixed VDI and compute workloads, and resource sharing across multiple VMs.
Additional resources
5.4. NVIDIA GPU features for Red Hat OpenShift Service on AWS
- NVIDIA Container Toolkit
- NVIDIA Container Toolkit enables you to create and run GPU-accelerated containers. The toolkit includes a container runtime library and utilities to automatically configure containers to use NVIDIA GPUs.
- NVIDIA AI Enterprise
NVIDIA AI Enterprise is an end-to-end, cloud-native suite of AI and data analytics software optimized, certified, and supported with NVIDIA-Certified systems.
NVIDIA AI Enterprise includes support for Red Hat Red Hat OpenShift Service on AWS. The following installation methods are supported:
- Red Hat OpenShift Service on AWS on bare metal or VMware vSphere with GPU Passthrough.
- Red Hat OpenShift Service on AWS on VMware vSphere with NVIDIA vGPU.
- GPU Feature Discovery
NVIDIA GPU Feature Discovery for Kubernetes is a software component that enables you to automatically generate labels for the GPUs available on a node. GPU Feature Discovery uses node feature discovery (NFD) to perform this labeling.
The Node Feature Discovery Operator (NFD) manages the discovery of hardware features and configurations in an OpenShift Container Platform cluster by labeling nodes with hardware-specific information. NFD labels the host with node-specific attributes, such as PCI cards, kernel, OS version, and so on.
You can find the NFD Operator in the Operator Hub by searching for “Node Feature Discovery”.
- NVIDIA GPU Operator with OpenShift Virtualization
Up until this point, the GPU Operator only provisioned worker nodes to run GPU-accelerated containers. Now, the GPU Operator can also be used to provision worker nodes for running GPU-accelerated virtual machines (VMs).
You can configure the GPU Operator to deploy different software components to worker nodes depending on which GPU workload is configured to run on those nodes.
- GPU Monitoring dashboard
- You can install a monitoring dashboard to display GPU usage information on the cluster Observe page in the Red Hat OpenShift Service on AWS web console. GPU utilization information includes the number of available GPUs, power consumption (in watts), temperature (in degrees Celsius), utilization (in percent), and other metrics for each GPU.
