Abstract

Enabling real-time processing of financial data streams is extremely challenging, especially considering that typical operations that interest investors often require combining data across (a potentially quadratic number of) different pairs of stocks. In this paper we present the architecture and the components of our system for the real-time processing of geo-distributed financial data at scale. Our system can scale to larger resources and utilizes a Synopses Data Engine in order to efficiently handle complex cross-stock queries, such as the ones required to detect systemic risk or to help forecast the value of some stock. The rich set of supported operations is depicted at the Visual Analytics component of our system.