DDN的动态

Enjoyed speaking with folks at the Lenovo booth this week at #Supercomputing2024 in Atlanta, discussing artificial intelligence in higher education, specifically around managing research data with AI. Think about this: How many research projects do you have at your higher education institution? For the research projects you run, how are you managing your research data??How complex is it??How efficient is it? More often than not, my friends researching topics to solve some of the world's biggest challenges run into data fragmentation, inefficient data access, compliance issues, and so much more. Enter AI. Here's the stack I'm thinking about: - Lenovo ThinkSystem SR665 or ThinkSystem SR950 servers, optimized for AI workloads - DDN Exascaler, a highly scalable, parallel file system offering fast access to large-scale datasets, suitable for AI workloads that require high-performance I/O, as well as DDN Infinia for real-time data management and insight capabilities, improving the management of unstructured data across the research lifecycle – NVIDIA H100 GPUs or GB200 GPUs are both great options here - NVIDIA Software: - CUDA, which is essential for developing AI applications well. - NIMS, accelerated inference microservices - check it out here: https://lnkd.in/gi3ryZg2 - Triton Inference Server to aid in AI model deployment - RAPIDS to accelerate data science pipelines with these GPUs - Middleware: Apache Kafka is a great option for real-time data streaming and integration to ensure data movement between manufacturing systems - Kubernetes to manage the AI-driven microservices and scaling resources dynamically - make optimal use of the containers in your solution - a number of options around data orchestration (e.g. Apache Airflow), AI/ML Libraries (e.g. PyTorch, TensorFlow), and data monitoring & observability (e.g. Prometheus, Grafana) What works in your environment? Looking for more open source or more commercial software? Have a different perspective? Let's talk - [email protected]. Dana Ryan, Scott Tease, Dave Mooney, Flynn Maloy, Ryan McCurdy, Ashley Gorakhpurwalla, Francois Corradino, Stuart McRae, David Ellison, Steve Eiland - BSEE