Idea:

Examine the queue state of a machine and report some statistics.  This
can then e.g. be collected on a single web page, giving an overview
over all available machines.

Data to be gathered:

- number of jobs in the queue, including number of own jobs
- number of queued, running, held jobs, including for own jobs
- requested node numbers and wall times of jobs, including for own jobs
- used run time of jobs, including for own jobs
- number of available, used, and free nodes
- if possible: estimated start time for a jobs of a given (which?) size
- remaining allocation, both SU and days left
- available disk space
- system info: number of nodes, processors per node, memory per node
- hardware info: processor type, operating system, GHz, MFlop/s
- some past usage statistics: jobs run, hours used, both own or of the
  whole group
- benchmark info: memory, network, and disk throughput and latency numbers