PRIMARY STORAGE

An append only data model is used to ensure job repeatability, so primary storage must be a location to which only trac has write-access. For cloud deployments this would be regular bucket storage.

EXECUTION SERVICE

An external compute service is invoked to execute jobs. For cloud deployments, we provide a Kubernetes pattern which supports both single-node and distributed (Spark) capabilities. Integrations with Hadoop and other compute services are available for on-premises deployments.

REPOSITORIES

Models are stored in an external repository, with the code being accessed dynamically at runtime. Integratations with GitHub, GitLab and Nexus are available as standard, custom repositories can also be accommodated using plug-ins.

NOTEBOOK / IDEs

Deploy the trac runtime library to your IDE or notebook to enjoy the type safety of production during the development process and build models which can be deployed to production with a single click and no code modification.

EXTERNAL STORAGE

Connecting to other data locations allows trac to manage the batch importing and exporting of data. Connectivity is available for several common technologies including; object storage, file-like locations, databases, and SQL-like locations.

AUTHENTICATION

Plug into your enterprise-wide SSO mechanism for watertight identity and access management, with roles assigned in the AD and fine-grain permissions being configured in trac. Azure AD, SAML and OpenID are supported out of the box, other solutions can be accomodated using plug-ins.