merge: Improve the OpenDC compute model (#37)

This pull request contains various improvements to the OpenDC compute simulation model.

- Support filtering hosts based on CPU capacity
- Do not allocate lambda in fast-path
- Redesign VM interference algorithm
- Report provisioning time of virtual machines
- Prevent allocations during collection cycle 
- Use correct flow input capacity for counters
- Support running workloads without coroutines

**Breaking API Changes**

- `VirtualMachine` now requires `cpuCapacity` parameter.
- `VmInterferenceModel` needs to be constructed using `VmInterferenceModel.Builder` and can't be passed a list of groups anymore.
- Scheduling latency is not collected anymore. Instead, use the boot time and provisioning time to derive the scheduling latency.
- Telemetry data is recorded using `*TableReader` interfaces as opposed to the `*Data` classes. These classes are re-used per row and should not be shared with other threads, since the underlying data may change.
- `SimMachine` does not implement `AutoCloseable` anymore. Machines can be removed from a `SimHypervisor` using the `removeMachine` method.
- `SimMachine.run` is moved to an extension method called `runWorkload`. Users can now also choose to run a workload using the asynchronous `SimMachine.startWorkload`.

0 files changed, 0 insertions, 0 deletions