Gordon and James; Happy to see this discussion. Gordon's scheduling ideas mirror the conversations we've had here concerning wide release of compute intensive next gen tools. This isn't much more than a me too e-mail but wanted to highlight the two points we also discussed.
2. Create 'classes' of tools: light/medium/heavy (or similar). 'light' tools are prioritized over 'heavy' tools (heavy usually means CPU bound).
The light/heavy distinction is the number one idea thrown around here. We have a lot of custom tools that fall into the heavy category and are memory intensive such that they can only run on certain hardware and only a certain number of jobs can run concurrently. Having them fall into a separate queues with different limits would help with this type of custom scheduling.
3. Limit on number of jobs-per-user.
This is also a point of contention due to the perception of potential processor hogs backing up jobs for days or weeks by submitting several intensive jobs. It's mostly a control issue to satisfy people worried about worst possible case situations. Brad