High Performance Computing by Hour

This service gives access to a virtual research environment with an expansive Earth Observation data archive. After a prototyping phase on a user VM, the Hadoop processing environment is readily available for upscaling larger datasets and to make large processing jobs more efficient. Prototyping on smaller datasets can be done using a virtual machine and can be upscaled using a couple of Spark code adaptations to the Hadoop cluster. Jobs on the Hadoop platform can be monitored using default yarn tools or a custom job-control application.

The amount of Hadoop cores (CPUs) is related to the available RAM, each Hadoop Core can use up to 3.5 GB RAM. If your processing is memory intensive, estimate the memory which is needed for your processing workflow and select the amount of cores accordingly. The hours in the hourly package are CPU hours when using the full capacity of the selected Hadoop Cores.

By selecting eg. 40 CPU cores, combined with the 100 hourly package, your processing has a maximal capacity of 40 CPU cores and up to 140GB of RAM. And this for 4000 (40 x 100) CPU hours of processing. If the process needs full capacity, it can process during 100 hours. However if it needs only 1 core, it can run for a total amount of 4000 processing hours.

The total amount of CPU hours will be available for your processing services during one calendar year after request. After this period of one year, the remaining CPU hours will be expired and no longer useable for your processing service.

To see all complementing products, please click here.