Core Principles

GPUs are acknowledged as the way forward in terms of executing large scale mathematical models over extremely large sets of data. ZettaMesh wholly embraces GPU computations as a core part of the platform and is designed from the ground up to be able to schedule operation on the GPU as needed. GPU acceleration allows ZettaMesh to perform feats such as currency conversion for a billion records in ~800ms and calculating 15,830,397 population intersections in ~1.5 seconds.

ZettaMesh is designed from the ground up for task parallelization without the need for explicit programming. ZettaMesh’s patent pending OpStream scheduling and execution engine is designed from the ground up to break down tasks into atomic execution units which can be scheduled to run in parallel on the most appropriate processing unit (GPU or CPU) for the task at hand; anywhere on the mesh. This allows ZettaMesh to intelligently and seamlessly schedule and execute tasks in parallel across the entire mesh. The ZettaMesh model authoring environment is designed to allow the end user to design the model without needing to think about parallelization, optimization or any other technology related concerns.

As implied in the name ZettaMesh is based around a revolutionary mesh architecture. In ZettaMesh’s infinitely expandable mesh architecture nodes can simply be added to the mesh to add processing capacity. The mesh architecture allows nodes to intelligently cooperate in executing parallel tasks using the ZettaMesh OpStream engine. Tasks are defined by the execution engine and nodes use a rule based mechanism to pick up those tasks based on logical considerations such as data and task locality, execution chains, etc….

ZettaMesh is designed from the ground up to be easily integrated into your existing big data solutions and workflows. ZettaMesh is designed to seamlessly integrate with Hadoop and other HDFS based solutions, web services, file based feeds, RDBMs and Excel file based data sources; thus allowing data from many different sources to be pulled into a ZettaMesh model.In addition to data integration ZettaMesh offers industry standard REST based web service APIs. These APIs allows ZettaMesh to be easily added into Spark flows or to be used in conjunction with any process management and orchestration solution which supports REST based integration.

One of the key core tenets of ZettaMesh is ease of use. ZettaMesh is designed from the ground up to allow the end user to easily and seamlessly define data models, mathematical models and analysis in an intuitive and easy to use fashion. The system handles all parallelization, data retrieval and management tasks, etc.… on behalf of the user; thus allowing the end user to focus on the models and analysis, not the technology required to execute them efficiently.