Many data analytics vendors are attempting to retrofit streaming technology into their conventional batch processing infrastructure. In contrast, the CeleraOne Analytics Engine was designed and built from the ground up for the most demanding streaming workloads. We pair horizontal scalability with great single-node performance to significantly reduce our customer's costs of operation.
In a presentation given at the Hadoop Get-Together Berlin, April 18, 2012, Dr. Falk-Florian Henrich explains how CeleraOne applies advanced compiler technology to speed up event stream processing.
1
Event Stream Processing
Event data continuously flows into the system as it is generated by users, customers, or machines. Stream processing as a mode of operation is prerequisite for obtaining results in near real-time.2
Full Sequence Analysis
In contrast to conventional CEP systems analysis of event streams is not limited to small data windows. Process long runs of event sequences and time series. Detect long-distance features and relationships.3
Large Scale Graph Processing
While processing events from thousands of different sources in parallel automatically build a time-dependent communications graph that reveals interactions between event sources.4
Auto-Parallelized Operations
Operations on data streams are specified in a simple SQL-like dialect. They are auto-parallelized and compiled to native code before execution. Queries execute as fast as hand-written C code.5
In-memory Technology
CeleraOne's unique in-memory technology provides for compression of event data and speed. Avoiding disk I/O altogether yields significant performance gains and ensures very low latencies.6
