Suggested new job visualization

An idea for a visualization to allow a user to view the progress of a job over time - here's a mockup I drew (badly!) on a whiteboard:


Time flows from left to right on the x-axis; All of the compute nodes are listed on the y-axis.

Tasks
* appear as lines, vertically aligned with the relevant compute node
* horizontally they trace their execution time
* colour coded by success/failure

Interactions
* Click on compute node to go to a detail page
* Click on a task to go to that task
* Zoom in to see details (perhaps show task ids within the bars?), pan left/right/up/down

What does this visualization show?

* Pool utilization - whitespace shows if the user getting maximum value from their pool, or is there a lot of idle time
* Problem nodes - if every task fails on a particular node (see #4 in the image above), that node might need checking.
* Overloads - if a tasks on a node take much longer to run than everywhere else, that node might have need checking
* Problem times - if every active task fails at the same time, might point to an external issue (throttling, storage, network connectivity, etc)

Issues to consider

* Should we show dependencies?
* How to handle jobs with thousands of tasks?