Details
-
Enabler
-
Not Assigned
-
None
-
Services
-
-
-
Inter Program
-
-
-
16.6
Description
As the epic progress the DP team will need to carry out tests of DP pipelines at significant computational scale.
Precise requirements are not known but basic modelling indicates DP tests will need to handle data sets up to of order 100 TB size, with networked storage I/O greater than 10GB/s by the end of 2023. In 2023 Q1 we probably need smaller system (e.g., ~10TB size storage, scaled...) and as initial tests progress better estimates of future needs will be known.
It is likely DP will need to test a range of machine configurations, e.g., large nodes with ~2TB RAM and 8 GPUS down to smaller single GPU nodes, possibly also just CPU nodes.
The duty cycle of the tests is expected to be low, perhaps a few hours / week, with the most demanding tests only carried out a few times.
Need to develop a plan for where, when and how these platforms will be accessible.
