Container Compute Service
We provide the best fully-managed AI computing environment in the network, with services including interactive containers and scheduled containers. The complete container management environment perfectly integrated with the storage space provides a convenient, simple, and easy-to-understand interface. Just clicks away from accessing a completely optimized AI computing environment of NGC.
Container Compute Service
Taiwan AI Cloud’s Container Compute Service provides two types of container environments, including interactive containers and scheduled containers, which can:
- Rapidly deploy development environment: Quickly edit programs via the built-in Jupyter Notebook, and choose from a variety of AI frameworks.
- Mount Taiwan AI Cloud’s Hyper File System (HFS) to containers automatically.
- Deploy and execute a large number of AI operations simultaneously to increase productivity.
Strengths
A great number of GPU resources meet high volume AI training requirements, reducing training time by nearly 50%, and doubling the inference performance.
Kubernetes architecture is adopted and NVIDIA is imported to optimize the AI software stack, rapidly deploy the working environment, and reduce container environment deployment time by 30%.
Large memory containers easily facilitate evaluations and conversion of specifications, and enable the deployment and execution of a large number of AI containers at the same time to increase productivity.
A secure and fast storage system that makes backup and deployment of the same work environment easy.
Features
- It provides an advanced NVIDIA® Tesla V100 32GB SMX GPU compute accelerator with NVLink interconnection under multiple GPUs.
- Equipped with at least 1 V100 GPU + 4 CPU entity cores + 90GB main memory, up to 8 GPUs can be called for parallel computing to accommodate large AI calculations.
- It provides images with various AI compute frameworks optimized by NVIDIA GPU Cloud, including common AI frameworks such as TensorFlow, Caffe, PyTorch, MXNet, Theano, and RAPIDS.
- Fully accessible with Jupyter Notebook for writing programs, online debugging, and computation.
- Scheduled Container can specify the starting time of the computation, with appropriate scheduling and planning of the quantity of resource for batch tasks.
- UI, API, CLI, and other tools are offered to manage containers.
- SSH or Jupyter Notebook can be used to connect to containers.
- The image feature allows easy replication of the container environment and the creation of multiple identical containers for computing.
- It is fully integrated with HFS fully-managed file system, and storage space is readily available.
Free Consultation Service
Contact our experts to learn more and get started with the best solution for you.