Alibaba Cloud heterogeneous computing services is a complete service system with integrated software and hardware that helps you allocate and scale resources in a flexible and elastic manner, increase computing power, and control costs.
Heterogeneous computing is a systematic computing method that consists of computing units of different instruction set and architecture types. Alibaba Cloud heterogeneous computing service family includes Elastic GPU Service, FPGA as a Service (FaaS), and Elastic Accelerated Computing Instances (EAIS). In heterogeneous computing, dedicated hardware is used to serve their most suitable scenarios. This way, in specific scenarios, heterogeneous computing instances can achieve efficiency and cost-effectiveness higher than those of common Elastic Compute Service (ECS) instances by one or more orders of magnitude. Heterogeneous computing is a technology that offers a balance of performance, cost, and power consumption to optimize performance and costs.
The rapid development of AI technologies such as deep learning has given birth to increasingly complex and accurate AI computing models and a significant increase in demand for computing power and performance. Against this backdrop, more and more AI computing services accelerate their performance by using heterogeneous computing. Cloud-based AI accelerators developed by Alibaba Cloud for heterogeneous computing services use a unified framework to accelerate major AI computing frameworks such as TensorFlow, PyTorch, MxNet, and Caffe and optimize the performance of Ethernet and heterogeneous accelerators.
Heterogeneous computing service family
This section describes the Alibaba Cloud heterogeneous computing service family, which consists of the following services: Elastic GPU Service, FaaS, Apsara AI Accelerator (AIACC), FastGPU, cGPU, and EAIS.
- Elastic GPU Service
GPU-accelerated instances are computing servers based on GPUs. GPUs have unique advantages over CPUs in mathematical and geometric computations such as floating-point and parallel computing and can provide 100 times the computing power of CPUs. GPU-accelerated instances combine the computing power of GPUs and CPUs and provide ready-to-use, scalable GPU compute resources for a variety of scenarios such as AI, high-performance computing, and professional graphics processing. For more information, see What is Elastic GPU Service?.
AIACC is an AI acceleration engine developed by Alibaba Cloud based on Infrastructure as a Service (IaaS) resources. It optimizes the models built on mainstream AI computing frameworks to achieve significant gains in training and inference performance in deep learning scenarios. You can use AIACC in conjunction with the cluster deployment tool FastGPU to build AI computing tasks to increase R&D efficiency and GPU utilization, reduce computing times, and lower latency in AI inference. For more information, see What is AIACC?.
FastGPU is a set of fast deployment tools provided by Alibaba Cloud for AI computing. You can use the interfaces and automatic tools of FastGPU to build GPU clusters within minutes and set up an efficient training environment for deep learning. For more information, see What is FastGPU?.
cGPU is a GPU-shared container technology developed by Alibaba Cloud. It provides kernel-based isolation of virtual GPU resources and allows you to deploy multiple containers on a single GPU. This way, you can better utilize your GPU hardware resources at lower costs while securely isolating your business. For more information, see What is the cGPU service?.
FPGA-accelerated instances are instances that are equipped with FPGAs. FPGA hardware and acceleration applications can be reconfigured to obtain low-latency hardware and elastic resources. For more information, see What are FPGA-accelerated instances?.