The Qube database is a point-in-time NoSQL database that is optimized for both time and cross-sectional queries and heavily compressed for backtest speed (datasets are on average 10 times smaller than in a SQL database).
It can store both external data (from data providers or internal sources) and calculated financial objects (the building blocks of an investment process) on frequencies ranging from yearly to intraday.
Each element possesses individual permissions. The data are organized into a unified, multi-asset-class referential so that data from different providers can readily be combined.
Our generic loader aims at describing any data source through a series of loading tasks so that it is feed and format agnostic.
The tasks are defined in xml, along with their interdependencies.
The generic loader schedules and resolves dependencies between tasks, so that no task is delayed unnecessarily. It warns the operator in case of a problem.
StarQube provides a rich set of APIs (C, C++, Java, C#, Matlab, Excel) to provide total freedom while relying on a common framework. Any financial object defined in the interface can be called using a simple syntax and FQL can be evaluated on the fly. Typical uses would be to feed Matlab with pre-arranged financial data or to create self-updating reports through the Excel API.
Organize investment data once and for all, save time on each new analysis
Download all data to a single location, avoid double subscriptions
Do not get locked-in with a single data provider
Data is centralized, avoiding reporting discrepancies and double subscriptions
Avoid data loading bottlenecks
Run calculations in FQL through the APIs and use external tools only where they shine