SAS^® EVENT STREAM PROCESSING

Built for speed

Enables high-volume processing of millions of events per second and low-latency response times using newly integrated ONNX Runtime for expanded support for GPU acceleration on CUDA and TensorRT supported platforms.

Ingests & consolidates streaming IoT sources

Lets you consume today's IoT sources – including cloud or edge on-site streams – and extend in the future with an extensive suite of prebuilt connectors and adapters.

Ready for real-time action 

Provides you with situational awareness via customizable alerts, notifications and updates so you can react appropriately to what's happening or predicted to happen.

Flexible, open modeling environment

Easily define, test and refine models in a low code environment using an intuitive visual interface. Data scientist-friendly coding environments include familiar Python developer (e.g., Jupyter notebook).

Complete multiphase analytics

Embed SAS at the edge, in the fog and in the cloud, cleansing and analyzing data at each streaming event phase. In-depth SAS models are portable to the stream and the edge. Supports algorithms, for example, to perform support vector data description, robust principal component analysis, random forest, gradient boosting and streaming regression analysis.

Unified project deployment & server management

SAS Event Stream Manager lets you construct and manage reusable deployment templates, easily add new auto-discover ESP servers in cloud, load and unload ESP projects, dynamically allocate resources, and automatically scale up and down for automated elasticity and monitoring for better resource management. When ESP servers are offline or unavailable, retry capabilities support redeploying models when connectivity resumes.

Cloud native

Compatible with cloud technologies, including Docker and Kubernetes, for large-scale, elastic, multitenant, distributed services. Ensures continuous, secure and stable event pattern detection through patented, instantaneous 1+N way failover, native failover, guaranteed delivery without persistence and other fault-tolerance functions for resilient, highly available event stream processing.

In-stream learning model windows

Uses different window types – Train, Score, Calculate, Model Supervisor, Model Reader – for different tasks (e.g., specifying data stream input sources, patterns of interest and derived output actions).

Image & video analytics

Process video and still image data, including image pre-processing and object detection and classification, by combining streaming analytics, video and image ingestion with SAS and third-party machine learning frameworks using ONNX model formats.

GitHub resources

Get SAS code examples, libraries and tools for developers and users.