diff options
| author | Fabian Mastenbroek <mail.fabianm@gmail.com> | 2022-10-21 22:32:05 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2022-10-21 22:32:05 +0200 |
| commit | fa7fdbb0126ea465130961dc37c4ef2d6feb36e9 (patch) | |
| tree | 9cd46dd7970870b78990d6c35e8e2759d7cf5a13 /opendc-experiments/opendc-experiments-workflow/src | |
| parent | 29beb50018cf2ad87b252c6c080f8c5de4600349 (diff) | |
| parent | 290e1fe14460d91e4703e55ac5f05dbe7b4505f7 (diff) | |
merge: Implement multi-flow stages in simulator (#110)
This pull request introduces the new `flow2` multi-flow simulator into the OpenDC codebase
and adjust all existing modules to make use of this new simulator.
The new simulator models flow as a network of components, which can each receive flow
from (potentially) multiple other components. In the previous simulator, the framework itself
supported only single flows between components and required re-implementation of many
components to support multiplexing flows.
Initial benchmarks show performance improvements in the range 2x–4x for large scale experiments
such as the Capelin benchmarks.
## Implementation Notes :hammer_and_pick:
* Add support for multi-flow stages
* Support flow transformations
* Add forwarding flow multiplexer
* Expose metrics on FlowMultiplexer
* Re-implement network sim using flow2
* Re-implement power sim using flow2
* Re-implement compute sim using flow2
* Optimize workload implementation of SimTrace
* Remove old flow simulator
* Add log4j-core dependency
## External Dependencies :four_leaf_clover:
* N/A
## Breaking API Changes :warning:
* Removal of the `org.opendc.simulator.flow` package. You should now use
the new flow simulator located in `org.opendc.simulator.flow2`.
* `PowerModel` interface is replaced by the `CpuPowerModel` interface.
* `PowerDriver` interface is replaced by the `SimPsu` and `SimPsuFactory` interfaces.
* Removal of `SimTraceWorkload`. Instead, create a workload from a `SimTrace` using
`createWorkload(offset)`.
* `ScalingGovernor` has been split in a `ScalingGovernor` and `ScalingGovernorFactory`.
* All modules in `opendc-simulator` are now written in Java. This means that default
parameters are not supported anymore for these modules.
Diffstat (limited to 'opendc-experiments/opendc-experiments-workflow/src')
| -rw-r--r-- | opendc-experiments/opendc-experiments-workflow/src/main/kotlin/org/opendc/experiments/workflow/TraceHelpers.kt | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/opendc-experiments/opendc-experiments-workflow/src/main/kotlin/org/opendc/experiments/workflow/TraceHelpers.kt b/opendc-experiments/opendc-experiments-workflow/src/main/kotlin/org/opendc/experiments/workflow/TraceHelpers.kt index 2ae69949..4dc3a775 100644 --- a/opendc-experiments/opendc-experiments-workflow/src/main/kotlin/org/opendc/experiments/workflow/TraceHelpers.kt +++ b/opendc-experiments/opendc-experiments-workflow/src/main/kotlin/org/opendc/experiments/workflow/TraceHelpers.kt @@ -74,7 +74,7 @@ public fun Trace.toJobs(): List<Job> { val submitTime = reader.getInstant(TASK_SUBMIT_TIME)!! val runtime = reader.getDuration(TASK_RUNTIME)!! val flops: Long = 4000 * runtime.seconds * grantedCpus - val workload = SimFlopsWorkload(flops) + val workload = SimFlopsWorkload(flops, 1.0) val task = Task( UUID(0L, id), "<unnamed>", |