A very simple implementation of a background queue/worker. The cool part: we're calling a proc. Since Maglev deals with real objects we can actually persist a proc, which you can't do in other implementations because you can't marshal a proc.

Open up 3 terminals. In the first run:

./producer.rb

And in the other 2 run:

./worker.rb

Each worker will start pulling "jobs" off the queue and do them.

When running 4 producers & 4 workers, I saw the following resources used:

gem: 4 instances, 2.1, 2.8, 4.2 & 7.3 MB respectively
magelv-ruby: 8 instances, ~310KB each
stoned: 1 instance, 32MB
topaz: 8 instances, ~24MB each
total: ~242MB