GOES-R Impact on NCEP Computing

GOES-R is scheduled to launch in 2016. It will produce raw data at a rate of 100 Mbps and Level 0/1/2+ data at a rate of 16 TB per day. Level 1b data is processed into useful observations of current conditions for the entire Western Hemisphere. These L1b observations and Level 2+ data are inputs to initialize multiple EMC analysis and forecast models. They will be at temporal and spatial resolutions far greater than what is available today and will require efficient handling of the data within NCEP architecture for analysis and forecast modeling systems. Harris has designed a High Performance Computing (HPC) environment for NOAA's GOES-R Ground System to ensure all data are available with 5 minute latency and a minimum Operational Availability of 0.9999. The Harris GOES-R Ground System is a Service Oriented Architecture (SOA) composed of services for Enterprise Management, Mission Management, Product Generation, and Product Distribution. Each of these services provides high reliability through redundancy of critical components and functions, ensuring there is no single point of failure in the system. Additionally, Product Generation and Product Distribution services are highly parallelized to achieve the required 5 minute latency. A high throughput, low latency Enterprise Service Bus (ESB) is used for message traffic between services, and a highly reliable, high throughput Data Fabric is used for data transfers between algorithms services. Data communications between all services are provided through a high speed, low latency 10 gigabit Ethernet network. The GOES-R system is just one of the new generation of observation systems supporting NCEP modeling and forecasting. These systems will produce larger volumes of data, at higher frequencies, with better operational availability than the current observation systems that supply data to the NCEP. In developing GOES-R, multiple HPC and high reliability technologies were merged to build an efficient, fast, and reliable system. This poster will present the features of the GOES-R architecture allowing it to efficiently and reliably produce data in an HPC environment.