Designing for graph-structured data from the start.

Last Significant Update:

2024-06-25

Status:

Draft

Comments to:

mensurae@jkoshy.net

Mensurae’s data would be representing information about the behavior of systems during a measurement.

  • The data could represent an event that spans multiple machines, e.g., when representing the flow of a remote procedure call.

  • There could be cause-and-effect relationships between specific items of data.

  • There could be before-and-after relationships between specific items of data that could be used to sort out clock skew.

  • While some kinds of events could be nested (e.g., call-frame entry and exit), in general, the relationships between items could form a graph (perhaps a directed graph if a time ordering is imposed on the relationships).

  • User analysis may add further “layers” of relationships to this data; in general these relations too could have graph structure.

It seems best to design for graph-structured data from the start.