Abacus team is ready to start the next big chunk of work. We want to remove the time-based metrics we now use to integrate with Cloud Foundry's app & service events and replace them with sampling.
Why?
Time-based metrics are complex, costly in terms of CPU and storage space. Clients have to renew time-based usage, which requires persistence on their side. We see tham as unreliable and error-prone.
How?
We are proposing a new Sampler component that will expose REST API for continuous events.
