Here are components of AI stack
Storage array - AI requires massive amounts of data storage. Companies will deploy storage arrays with petabytes of storage, often using technologies like SAN/NAS storage networks. Popular options include Dell EMC, NetApp, Pure Storage, etc.
•High performance computing - AI models require enormous processing power. Companies will deploy clusters of high-end GPUs and CPUs. Popular options include NVIDIA DGX servers, IBM Power servers, etc.
•Networking equipment - Fast networking equipment is needed to connect the storage arrays, compute clusters, and other infrastructure. Options include InfiniBand or 100GbE networking platforms. Cisco, Mellanox, and Arista Networks are popular vendors.
•Orchestration software - Software is needed to orchestrate jobs across the infrastructure. Popular options include Kubernetes, OpenStack, AWS Batch, etc. These help deploy containers and manage clusters.
•Cloud services - Companies will often use cloud platforms like AWS, Azure and GCP to help scale their AI computing needs. The cloud provides on-demand access to storage, computing, and many AI services.
•Visualization and workstations - Data scientists and engineers need powerful workstations to access the infrastructure, build models, visualize data, etc. These often include advanced GPUs, high resolution monitors, lots of RAM and fast I/O.
•Software libraries and frameworks - The environment needs to have many software libraries installed to support AI development like TensorFlow, PyTorch, Scikit-Learn, OpenCV, etc. These provide the tools to build, train and deploy models.
•Management software - Software is also needed to help monitor, manage and optimize the entire AI computing environment. Things like workload managers, monitoring dashboards, log aggregators, etc.
•Security - Robust security solutions are required to secure data, infrastructure and workloads in the environment. Things like identity access management, encryption, firewalls, etc.