Teradata is a message passing system. Messages are sent from parsing engines to AMPs, and from AMPs to AMPs, and from AMPs to parsing engines. That’s the key way that components in a shared nothing architecture pass data and work requests among themselves.
When a message arrives on an AMP and the message represents work that needs to get done on the AMP, that message is assigned a“work type”, depending on the importance of the work-to-be-done. There are 16 different work types supported: Work00 to Work15.
Under usual conditions, all load utility jobs and all queries run using AMP worker tasks (AWTs) of the same message work types: Work00, Work01, and Work02. However, if you increase AWTs per AMP above a certain threshold, then all your utility jobs will be assigned to different work types and given their own reserve pools.
If you are someone who monitors or is otherwise interested in AWTs and how they are being used, this posting describes changes related to your utility jobs, and what options you have for managing these changes.