The key bottleneck in the current implementation is the decoding of JSON and then the analytics happening inside ReFrame. We should update the storage schema and avoid the in-Python decoding of JSON in favor of direct DB queries. Preliminary experiments show that this is considerably faster.
The key bottleneck in the current implementation is the decoding of JSON and then the analytics happening inside ReFrame. We should update the storage schema and avoid the in-Python decoding of JSON in favor of direct DB queries. Preliminary experiments show that this is considerably faster.