Fine-grained access control to Ray clusters/workloads. Using AWS Identity and Access Management (IAM)
In-transit & at-rest encryption. Using AWS Certificate Manager (ACM) & AWS Key Management Service (AWS KMS)
Object Storage & distributed file access. Using Amazon Simple Storage Service (Amazon S3) & Amazon Elastic File System (Amazon EFS)
Observability and metrics. Using Amazon CloudWatch.
Wildlife Studios runs Anyscale on AWS to serve in-game offers 3x faster and at 10% of the cost, with the potential for $400,000 in infrastructure savings annually.
Dendra Systems runs Anyscale on AWS to do massive-scale machine learning on aerial imagery for environmental restoration and planting trees. This involves terabyte-scale processing of image datasets and scaling deep neural network training with the Ray ecosystem. Read the blog.
Switching to running their machine learning workloads with Ray and Anyscale on AWS enabled Anastasia to accelerate their machine learning job completion times by 9x. Furthermore, they reduced the cost of running these workloads by over 80%.
Serverless computing is an important direction in the evolution of computing. Marvin Theimer, a distinguished engineer at Amazon speaks about the lessons learned in this domain.
Amazon uses Ray to process petabytes of data every single day. This has led to more than an order of magnitude improvement in scalability, throughput, and cost reduction.
Nixtla uses Ray on AWS to scale forecasting by fitting one million time series models in less than half an hour and for under $30.
Learn how AWS SageMaker and Ray are used together to manage reinforcement larning workflows from perception to controls to optimization and how this combination is used to create end-to-end solutions for power plant and manufacturing plant operations.