HPC and HTC Application Execution and Data Management

  • Unified API for scheduling jobs to run on a variety of remote resources, including supercomputers, Kubernetes clusters, physical servers and Virtual Machines.
  • Automates the data management lifecycle associated with a job including staging data inputs to the execution target and archive job outputs to storage resources.
  • Leverage containerized application assets to enable portability, and reduce the overall time-to-solution by utilizing data locality and other “smart scheduling” techniques.

Functions-as-a-service, Sensor Data/Streaming Data API & Events-Driven Workloads

  • Provides a distributed computing platform where computational primitives are based on Docker container images; containers are executed on Tapis cloud infrastructure in response to messages sent over HTTP.
  • Trigger executions based on Tapis events (e.g., file uploads or data streams); automatically scale functions to run in parallel as message load increases.
  • “Serverless” computing model - No servers to manage for end user.
  • Storing and retrieving sensor data for batch job processing, with support for temporal and spatial indexes and queries.
  • Automated, event-driven data stream processing workflows with integration into Tapis functions
  • Automated data management and scheduled archiving based on programmable policies.

Highly Scalable Document Store and Metadata API

  • Store and scale research data collections to billions of documents serialized using popular data formats such as JSON.
  • Configure custom indexes to optimize performance for specific usage patterns.

Identity, Authorization and Federated Security

  • Federated, decentralized model where each site/institution can manage the credentials and other secrets needed to access their compute resources.
  • Robust authorization based on scalable permissions, groups and roles.
  • Pluguble identity provider to leverage local or regional identities or a federated provider such as InCommon.
  • Secure authentication with short-term tokens based on OAuth2. Containers, Reproducibility & Smart Scheduling

Containers, Reproducibility & Smart Scheduling

  • The API remembers inputs and parameters used for each job so that computations can be repeated.
  • The API tracks which users modified which assets (files, apps, actors, permissions, etc)
  • Supports gathering usage metrics for reporting to funding agencies as well as security incident analysis.