Vector Execution Model

The dbX software stack has been purpose developed to leverage the technology of today and tomorrow. The core engine uses a vector execution model to maximize the payload at every stage of the processing and minimize the overheads. This model is fundamentally different from the row-at-a-time execution model of traditional database engines. Legacy engines, open-source as well as proprietary, were developed during computing eras where CPU cycles, memory and network bandwidth were all in short supply. These constraints drove the architecture of the execution engines towards operating on a few rows at a time and minimizing Network traffic. 

Since these hardware constraints no longer apply, we have adopted the best architecture that can leverage current technology. The dbX software stack is highly multi-threaded to use as many CPU cores are available. The software operates on large vectors of rows and all external I/O to Disks or Network is performed via large memory buffers to minimize overheads and maximize throughput.

Legacy database engines make very inefficient use of current hardware resources. These inefficiencies are masked when data volumes are small, but get magnified for Big Data.