Merge pull request #69 from atlarge-research/refactor/workflows-v1

Refactor workflow service to schedule tasks onto VMs

4 files changed, 528 insertions, 484 deletions

diff --git a/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimBareMetalMachine.kt b/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimBareMetalMachine.kt
index 4340708f..5e50a676 100644
--- a/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimBareMetalMachine.kt
+++ b/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimBareMetalMachine.kt
@@ -251,7 +251,7 @@ public class SimBareMetalMachine(
 
             this.isEmpty = !nonEmpty
             this.totalUsage = totalUsage
-            this.minExit = minExit
+            this.minExit = if (isEmpty) 0 else minExit
             this.maxExit = maxExit
         }
 
diff --git a/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimFairSharedHypervisor.kt b/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimFairSharedHypervisor.kt
new file mode 100644
index 00000000..b88871a5
--- /dev/null
+++ b/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimFairSharedHypervisor.kt
@@ -0,0 +1,517 @@
+/*
+ * Copyright (c) 2020 AtLarge Research
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+package org.opendc.simulator.compute
+
+import kotlinx.coroutines.*
+import kotlinx.coroutines.channels.Channel
+import kotlinx.coroutines.flow.MutableStateFlow
+import kotlinx.coroutines.flow.StateFlow
+import kotlinx.coroutines.intrinsics.startCoroutineCancellable
+import kotlinx.coroutines.selects.SelectClause0
+import kotlinx.coroutines.selects.SelectInstance
+import kotlinx.coroutines.selects.select
+import org.opendc.simulator.compute.interference.PerformanceInterferenceModel
+import org.opendc.simulator.compute.model.ProcessingUnit
+import org.opendc.simulator.compute.workload.SimWorkload
+import java.time.Clock
+import kotlin.math.ceil
+import kotlin.math.max
+import kotlin.math.min
+
+/**
+ * A [SimHypervisor] that distributes the computing requirements of multiple [SimWorkload] on a single
+ * [SimBareMetalMachine] concurrently using weighted fair sharing.
+ *
+ * @param coroutineScope The [CoroutineScope] to run the simulated workloads in.
+ * @param clock The virtual clock to track the simulation time.
+ */
+@OptIn(ExperimentalCoroutinesApi::class, InternalCoroutinesApi::class)
+public class SimFairSharedHypervisor(
+    private val coroutineScope: CoroutineScope,
+    private val clock: Clock,
+    private val listener: SimHypervisor.Listener? = null
+) : SimHypervisor {
+    /**
+     * A flag to indicate the driver is stopped.
+     */
+    private var stopped: Boolean = false
+
+    /**
+     * The channel for scheduling new CPU requests.
+     */
+    private val schedulingQueue = Channel<SchedulerCommand>(Channel.UNLIMITED)
+
+    /**
+     * Create a [SimMachine] instance on which users may run a [SimWorkload].
+     *
+     * @param model The machine to create.
+     */
+    override fun createMachine(model: SimMachineModel, performanceInterferenceModel: PerformanceInterferenceModel?): SimMachine {
+        val vm = VmSession(model, performanceInterferenceModel)
+        val vmCtx = VmExecutionContext(vm)
+
+        return object : SimMachine {
+            override val model: SimMachineModel
+                get() = vmCtx.machine
+
+            override val usage: StateFlow<Double>
+                get() = vm.usage
+
+            /**
+             * The current active workload.
+             */
+            private var activeWorkload: SimWorkload? = null
+
+            override suspend fun run(workload: SimWorkload) {
+                require(activeWorkload == null) { "Run should not be called concurrently" }
+
+                try {
+                    activeWorkload = workload
+                    workload.run(vmCtx)
+                } finally {
+                    activeWorkload = null
+                }
+            }
+
+            override fun toString(): String = "SimVirtualMachine"
+        }
+    }
+
+    /**
+     * Run the scheduling process of the hypervisor.
+     */
+    override suspend fun run(ctx: SimExecutionContext) {
+        val model = ctx.machine
+        val maxUsage = model.cpus.sumByDouble { it.frequency }
+        val pCPUs = model.cpus.indices.sortedBy { model.cpus[it].frequency }
+
+        val vms = mutableSetOf<VmSession>()
+        val vcpus = mutableListOf<VCpu>()
+
+        val usage = DoubleArray(model.cpus.size)
+        val burst = LongArray(model.cpus.size)
+
+        fun process(command: SchedulerCommand) {
+            when (command) {
+                is SchedulerCommand.Schedule -> {
+                    vms += command.vm
+                    vcpus.addAll(command.vm.vcpus)
+                }
+                is SchedulerCommand.Deschedule -> {
+                    vms -= command.vm
+                    vcpus.removeAll(command.vm.vcpus)
+                }
+                is SchedulerCommand.Interrupt -> {
+                }
+            }
+        }
+
+        fun processRemaining() {
+            var command = schedulingQueue.poll()
+            while (command != null) {
+                process(command)
+                command = schedulingQueue.poll()
+            }
+        }
+
+        while (!stopped) {
+            // Wait for a request to be submitted if we have no work yet.
+            if (vcpus.isEmpty()) {
+                process(schedulingQueue.receive())
+            }
+
+            processRemaining()
+
+            val start = clock.millis()
+
+            var duration: Double = Double.POSITIVE_INFINITY
+            var deadline: Long = Long.MAX_VALUE
+            var availableUsage = maxUsage
+            var totalRequestedUsage = 0.0
+            var totalRequestedBurst = 0L
+
+            // Sort the vCPUs based on their requested usage
+            // Profiling shows that it is faster to sort every slice instead of maintaining some kind of sorted set
+            vcpus.sort()
+
+            // Divide the available host capacity fairly across the vCPUs using max-min fair sharing
+            for ((i, req) in vcpus.withIndex()) {
+                val remaining = vcpus.size - i
+                val availableShare = availableUsage / remaining
+                val grantedUsage = min(req.limit, availableShare)
+
+                // Take into account the minimum deadline of this slice before we possible continue
+                deadline = min(deadline, req.vm.deadline)
+
+                // Ignore empty CPUs
+                if (grantedUsage <= 0 || req.burst <= 0) {
+                    req.allocatedLimit = 0.0
+                    continue
+                }
+
+                totalRequestedUsage += req.limit
+                totalRequestedBurst += req.burst
+
+                req.allocatedLimit = grantedUsage
+                availableUsage -= grantedUsage
+
+                // The duration that we want to run is that of the shortest request from a vCPU
+                duration = min(duration, req.burst / grantedUsage)
+            }
+
+            val totalAllocatedUsage = maxUsage - availableUsage
+            var totalAllocatedBurst = 0L
+            availableUsage = totalAllocatedUsage
+            val serverLoad = totalAllocatedUsage / maxUsage
+
+            // / XXX Ceil duration to eliminate rounding issues
+            duration = ceil(duration)
+
+            // Divide the requests over the available capacity of the pCPUs fairly
+            for (i in pCPUs) {
+                val maxCpuUsage = model.cpus[i].frequency
+                val fraction = maxCpuUsage / maxUsage
+                val grantedUsage = min(maxCpuUsage, totalAllocatedUsage * fraction)
+                val grantedBurst = ceil(duration * grantedUsage).toLong()
+
+                usage[i] = grantedUsage
+                burst[i] = grantedBurst
+                totalAllocatedBurst += grantedBurst
+                availableUsage -= grantedUsage
+            }
+
+            // XXX If none of the VMs require any computation, wait until their deadline, otherwise trigger on the
+            // first vCPU finished.
+            val triggerMode =
+                if (totalAllocatedBurst > 0 && totalAllocatedUsage > 0.0)
+                    SimExecutionContext.TriggerMode.FIRST
+                else
+                    SimExecutionContext.TriggerMode.DEADLINE
+
+            // We run the total burst on the host processor. Note that this call may be cancelled at any moment in
+            // time, so not all of the burst may be executed.
+            val isInterrupted = select<Boolean> {
+                schedulingQueue.onReceive { schedulingQueue.offer(it); true }
+                ctx.onRun(SimExecutionContext.Slice(burst, usage, deadline), triggerMode)
+                    .invoke { false }
+            }
+
+            val end = clock.millis()
+
+            // The total requested burst that the VMs wanted to run in the time-frame that we ran.
+            val totalRequestedSubBurst =
+                vcpus.map { ceil((duration * 1000) / (it.vm.deadline - start) * it.burst).toLong() }.sum()
+            val totalRemainder = burst.sum()
+            val totalGrantedBurst = totalAllocatedBurst - totalRemainder
+
+            // The burst that was lost due to overcommissioning of CPU resources
+            var totalOvercommissionedBurst = 0L
+            // The burst that was lost due to interference.
+            var totalInterferedBurst = 0L
+
+            val vmIterator = vms.iterator()
+            while (vmIterator.hasNext()) {
+                val vm = vmIterator.next()
+
+                // Apply performance interference model
+                val performanceScore = vm.performanceInterferenceModel?.apply(serverLoad) ?: 1.0
+                var hasFinished = false
+
+                for (vcpu in vm.vcpus) {
+                    // Compute the fraction of compute time allocated to the VM
+                    val fraction = vcpu.allocatedLimit / totalAllocatedUsage
+
+                    // Compute the burst time that the VM was actually granted
+                    val grantedBurst = ceil(totalGrantedBurst * fraction).toLong()
+
+                    // The burst that was actually used by the VM
+                    val usedBurst = ceil(grantedBurst * performanceScore).toLong()
+
+                    totalInterferedBurst += grantedBurst - usedBurst
+
+                    // Compute remaining burst time to be executed for the request
+                    if (vcpu.consume(usedBurst)) {
+                        hasFinished = true
+                    } else if (vm.deadline <= end) {
+                        // Request must have its entire burst consumed or otherwise we have overcommission
+                        // Note that we count the overcommissioned burst if the hypervisor has failed.
+                        totalOvercommissionedBurst += vcpu.burst
+                    }
+                }
+
+                if (hasFinished || vm.deadline <= end) {
+                    // Mark the VM as finished and deschedule the VMs if needed
+                    if (vm.finish()) {
+                        vmIterator.remove()
+                        vcpus.removeAll(vm.vcpus)
+                    }
+                }
+            }
+
+            listener?.onSliceFinish(
+                this,
+                totalRequestedBurst,
+                min(totalRequestedSubBurst, totalGrantedBurst), // We can run more than requested due to timing
+                totalOvercommissionedBurst,
+                totalInterferedBurst, // Might be smaller than zero due to FP rounding errors,
+                min(
+                    totalAllocatedUsage,
+                    totalRequestedUsage
+                ), // The allocated usage might be slightly higher due to FP rounding
+                totalRequestedUsage
+            )
+        }
+    }
+
+    /**
+     * A scheduling command processed by the scheduler.
+     */
+    private sealed class SchedulerCommand {
+        /**
+         * Schedule the specified VM on the hypervisor.
+         */
+        data class Schedule(val vm: VmSession) : SchedulerCommand()
+
+        /**
+         * De-schedule the specified VM on the hypervisor.
+         */
+        data class Deschedule(val vm: VmSession) : SchedulerCommand()
+
+        /**
+         * Interrupt the scheduler.
+         */
+        object Interrupt : SchedulerCommand()
+    }
+
+    /**
+     * A virtual machine running on the hypervisor.
+     *
+     * @param ctx The execution context the vCPU runs in.
+     * @param triggerMode The mode when to trigger the VM exit.
+     * @param merge The function to merge consecutive slices on spillover.
+     * @param select The function to select on finish.
+     */
+    @OptIn(InternalCoroutinesApi::class)
+    private data class VmSession(
+        val model: SimMachineModel,
+        val performanceInterferenceModel: PerformanceInterferenceModel? = null,
+        var triggerMode: SimExecutionContext.TriggerMode = SimExecutionContext.TriggerMode.FIRST,
+        var merge: (SimExecutionContext.Slice, SimExecutionContext.Slice) -> SimExecutionContext.Slice = { _, r -> r },
+        var select: () -> Unit = {}
+    ) {
+        /**
+         * The vCPUs of this virtual machine.
+         */
+        val vcpus: List<VCpu>
+
+        /**
+         * The slices that the VM wants to run.
+         */
+        var queue: Iterator<SimExecutionContext.Slice> = emptyList<SimExecutionContext.Slice>().iterator()
+
+        /**
+         * The current active slice.
+         */
+        var activeSlice: SimExecutionContext.Slice? = null
+
+        /**
+         * The current deadline of the VM.
+         */
+        val deadline: Long
+            get() = activeSlice?.deadline ?: Long.MAX_VALUE
+
+        /**
+         * A flag to indicate that the VM is idle.
+         */
+        val isIdle: Boolean
+            get() = activeSlice == null
+
+        /**
+         * The usage of the virtual machine.
+         */
+        val usage: MutableStateFlow<Double> = MutableStateFlow(0.0)
+
+        init {
+            vcpus = model.cpus.mapIndexed { i, model -> VCpu(this, model, i) }
+        }
+
+        /**
+         * Schedule the given slices on this vCPU, replacing the existing slices.
+         */
+        fun schedule(slices: Sequence<SimExecutionContext.Slice>) {
+            queue = slices.iterator()
+
+            if (queue.hasNext()) {
+                activeSlice = queue.next()
+                refresh()
+            }
+        }
+
+        /**
+         * Cancel the existing workload on the VM.
+         */
+        fun cancel() {
+            queue = emptyList<SimExecutionContext.Slice>().iterator()
+            activeSlice = null
+            refresh()
+        }
+
+        /**
+         * Finish the current slice of the VM.
+         *
+         * @return `true` if the vCPUs may be descheduled, `false` otherwise.
+         */
+        fun finish(): Boolean {
+            val activeSlice = activeSlice ?: return true
+
+            return if (queue.hasNext()) {
+                val needsMerge = activeSlice.burst.any { it > 0 }
+                val candidateSlice = queue.next()
+                val slice = if (needsMerge) merge(activeSlice, candidateSlice) else candidateSlice
+
+                this.activeSlice = slice
+
+                // Update the vCPU cache
+                refresh()
+
+                false
+            } else {
+                this.activeSlice = null
+                select()
+                true
+            }
+        }
+
+        /**
+         * Refresh the vCPU cache.
+         */
+        fun refresh() {
+            vcpus.forEach { it.refresh() }
+            usage.value = vcpus.sumByDouble { it.burst / it.limit } / vcpus.size
+        }
+    }
+
+    /**
+     * A virtual CPU that can be scheduled on a physical CPU.
+     *
+     * @param vm The VM of which this vCPU is part.
+     * @param model The model of CPU that this vCPU models.
+     * @param id The id of the vCPU with respect to the VM.
+     */
+    private data class VCpu(
+        val vm: VmSession,
+        val model: ProcessingUnit,
+        val id: Int
+    ) : Comparable<VCpu> {
+        /**
+         * The current limit on the vCPU.
+         */
+        var limit: Double = 0.0
+
+        /**
+         * The limit allocated by the hypervisor.
+         */
+        var allocatedLimit: Double = 0.0
+
+        /**
+         * The current burst running on the vCPU.
+         */
+        var burst: Long = 0L
+
+        /**
+         * Consume the specified burst on this vCPU.
+         */
+        fun consume(burst: Long): Boolean {
+            this.burst = max(0, this.burst - burst)
+
+            // Flush the result to the slice if it exists
+            vm.activeSlice?.burst?.takeIf { id < it.size }?.set(id, this.burst)
+
+            val actuallyExists = vm.activeSlice?.burst?.let { id < it.size } ?: false
+            return actuallyExists && this.burst == 0L
+        }
+
+        /**
+         * Refresh the information of this vCPU based on the current slice.
+         */
+        fun refresh() {
+            limit = vm.activeSlice?.limit?.takeIf { id < it.size }?.get(id) ?: 0.0
+            burst = vm.activeSlice?.burst?.takeIf { id < it.size }?.get(id) ?: 0
+        }
+
+        /**
+         * Compare to another vCPU based on the current load of the vCPU.
+         */
+        override fun compareTo(other: VCpu): Int {
+            return limit.compareTo(other.limit)
+        }
+
+        /**
+         * Create a string representation of the vCPU.
+         */
+        override fun toString(): String =
+            "vCPU(id=$id,burst=$burst,limit=$limit,allocatedLimit=$allocatedLimit)"
+    }
+
+    /**
+     * The execution context in which a VM runs.
+     *
+     */
+    private inner class VmExecutionContext(val session: VmSession) :
+        SimExecutionContext, DisposableHandle {
+        override val machine: SimMachineModel
+            get() = session.model
+
+        override val clock: Clock
+            get() = this@SimFairSharedHypervisor.clock
+
+        @OptIn(InternalCoroutinesApi::class)
+        override fun onRun(
+            batch: Sequence<SimExecutionContext.Slice>,
+            triggerMode: SimExecutionContext.TriggerMode,
+            merge: (SimExecutionContext.Slice, SimExecutionContext.Slice) -> SimExecutionContext.Slice
+        ): SelectClause0 = object : SelectClause0 {
+            @InternalCoroutinesApi
+            override fun <R> registerSelectClause0(select: SelectInstance<R>, block: suspend () -> R) {
+                session.triggerMode = triggerMode
+                session.merge = merge
+                session.select = {
+                    if (select.trySelect()) {
+                        block.startCoroutineCancellable(select.completion)
+                    }
+                }
+                session.schedule(batch)
+                // Indicate to the hypervisor that the VM should be re-scheduled
+                schedulingQueue.offer(SchedulerCommand.Schedule(session))
+                select.disposeOnSelect(this@VmExecutionContext)
+            }
+        }
+
+        override fun dispose() {
+            if (!session.isIdle) {
+                session.cancel()
+                schedulingQueue.offer(SchedulerCommand.Deschedule(session))
+            }
+        }
+    }
+}
diff --git a/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimHypervisor.kt b/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimHypervisor.kt
index 6087227b..fb4cd137 100644
--- a/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimHypervisor.kt
+++ b/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/SimHypervisor.kt
@@ -22,267 +22,23 @@
 
 package org.opendc.simulator.compute
 
-import kotlinx.coroutines.*
-import kotlinx.coroutines.channels.Channel
-import kotlinx.coroutines.flow.MutableStateFlow
-import kotlinx.coroutines.flow.StateFlow
-import kotlinx.coroutines.intrinsics.startCoroutineCancellable
-import kotlinx.coroutines.selects.SelectClause0
-import kotlinx.coroutines.selects.SelectInstance
-import kotlinx.coroutines.selects.select
 import org.opendc.simulator.compute.interference.PerformanceInterferenceModel
-import org.opendc.simulator.compute.model.ProcessingUnit
 import org.opendc.simulator.compute.workload.SimWorkload
-import java.time.Clock
-import kotlin.math.ceil
-import kotlin.math.max
-import kotlin.math.min
 
 /**
- * SimHypervisor distributes the computing requirements of multiple [SimWorkload] on a single [SimBareMetalMachine] concurrently.
- *
- * @param coroutineScope The [CoroutineScope] to run the simulated workloads in.
- * @param clock The virtual clock to track the simulation time.
+ * SimHypervisor distributes the computing requirements of multiple [SimWorkload] on a single [SimBareMetalMachine] i
+ * concurrently.
  */
-@OptIn(ExperimentalCoroutinesApi::class, InternalCoroutinesApi::class)
-public class SimHypervisor(
-    private val coroutineScope: CoroutineScope,
-    private val clock: Clock,
-    private val listener: Listener? = null
-) : SimWorkload {
-    /**
-     * A set for tracking the VM context objects.
-     */
-    private val vms: MutableSet<VmExecutionContext> = mutableSetOf()
-
-    /**
-     * A flag to indicate the driver is stopped.
-     */
-    private var stopped: Boolean = false
-
-    /**
-     * The channel for scheduling new CPU requests.
-     */
-    private val schedulingQueue = Channel<SchedulerCommand>(Channel.UNLIMITED)
-
+public interface SimHypervisor : SimWorkload {
     /**
      * Create a [SimMachine] instance on which users may run a [SimWorkload].
      *
      * @param model The machine to create.
      */
-    public fun createMachine(model: SimMachineModel, performanceInterferenceModel: PerformanceInterferenceModel? = null): SimMachine {
-        val vm = VmSession(model, performanceInterferenceModel)
-        val vmCtx = VmExecutionContext(vm)
-
-        return object : SimMachine {
-            override val model: SimMachineModel
-                get() = vmCtx.machine
-
-            override val usage: StateFlow<Double>
-                get() = vm.usage
-
-            /**
-             * The current active workload.
-             */
-            private var activeWorkload: SimWorkload? = null
-
-            override suspend fun run(workload: SimWorkload) {
-                require(activeWorkload == null) { "Run should not be called concurrently" }
-
-                try {
-                    activeWorkload = workload
-                    workload.run(vmCtx)
-                } finally {
-                    activeWorkload = null
-                }
-            }
-
-            override fun toString(): String = "SimVirtualMachine"
-        }
-    }
-
-    /**
-     * Run the scheduling process of the hypervisor.
-     */
-    override suspend fun run(ctx: SimExecutionContext) {
-        val model = ctx.machine
-        val maxUsage = model.cpus.sumByDouble { it.frequency }
-        val pCPUs = model.cpus.indices.sortedBy { model.cpus[it].frequency }
-
-        val vms = mutableSetOf<VmSession>()
-        val vcpus = mutableListOf<VCpu>()
-
-        val usage = DoubleArray(model.cpus.size)
-        val burst = LongArray(model.cpus.size)
-
-        fun process(command: SchedulerCommand) {
-            when (command) {
-                is SchedulerCommand.Schedule -> {
-                    vms += command.vm
-                    vcpus.addAll(command.vm.vcpus)
-                }
-                is SchedulerCommand.Deschedule -> {
-                    vms -= command.vm
-                    vcpus.removeAll(command.vm.vcpus)
-                }
-                is SchedulerCommand.Interrupt -> {
-                }
-            }
-        }
-
-        fun processRemaining() {
-            var command = schedulingQueue.poll()
-            while (command != null) {
-                process(command)
-                command = schedulingQueue.poll()
-            }
-        }
-
-        while (!stopped) {
-            // Wait for a request to be submitted if we have no work yet.
-            if (vcpus.isEmpty()) {
-                process(schedulingQueue.receive())
-            }
-
-            processRemaining()
-
-            val start = clock.millis()
-
-            var duration: Double = Double.POSITIVE_INFINITY
-            var deadline: Long = Long.MAX_VALUE
-            var availableUsage = maxUsage
-            var totalRequestedUsage = 0.0
-            var totalRequestedBurst = 0L
-
-            // Sort the vCPUs based on their requested usage
-            // Profiling shows that it is faster to sort every slice instead of maintaining some kind of sorted set
-            vcpus.sort()
-
-            // Divide the available host capacity fairly across the vCPUs using max-min fair sharing
-            for ((i, req) in vcpus.withIndex()) {
-                val remaining = vcpus.size - i
-                val availableShare = availableUsage / remaining
-                val grantedUsage = min(req.limit, availableShare)
-
-                // Take into account the minimum deadline of this slice before we possible continue
-                deadline = min(deadline, req.vm.deadline)
-
-                // Ignore empty CPUs
-                if (grantedUsage <= 0 || req.burst <= 0) {
-                    req.allocatedLimit = 0.0
-                    continue
-                }
-
-                totalRequestedUsage += req.limit
-                totalRequestedBurst += req.burst
-
-                req.allocatedLimit = grantedUsage
-                availableUsage -= grantedUsage
-
-                // The duration that we want to run is that of the shortest request from a vCPU
-                duration = min(duration, req.burst / grantedUsage)
-            }
-
-            // XXX We set the minimum duration to 5 minutes here to prevent the rounding issues that are occurring with the FLOPs.
-            duration = 300.0
-
-            val totalAllocatedUsage = maxUsage - availableUsage
-            var totalAllocatedBurst = 0L
-            availableUsage = totalAllocatedUsage
-            val serverLoad = totalAllocatedUsage / maxUsage
-
-            // Divide the requests over the available capacity of the pCPUs fairly
-            for (i in pCPUs) {
-                val maxCpuUsage = model.cpus[i].frequency
-                val fraction = maxCpuUsage / maxUsage
-                val grantedUsage = min(maxCpuUsage, totalAllocatedUsage * fraction)
-                val grantedBurst = ceil(duration * grantedUsage).toLong()
-
-                usage[i] = grantedUsage
-                burst[i] = grantedBurst
-                totalAllocatedBurst += grantedBurst
-                availableUsage -= grantedUsage
-            }
-
-            // We run the total burst on the host processor. Note that this call may be cancelled at any moment in
-            // time, so not all of the burst may be executed.
-            select<Boolean> {
-                schedulingQueue.onReceive { schedulingQueue.offer(it); true }
-                ctx.onRun(SimExecutionContext.Slice(burst, usage, deadline), SimExecutionContext.TriggerMode.DEADLINE)
-                    .invoke { false }
-            }
-
-            val end = clock.millis()
-
-            // No work was performed
-            if ((end - start) <= 0) {
-                continue
-            }
-
-            // The total requested burst that the VMs wanted to run in the time-frame that we ran.
-            val totalRequestedSubBurst =
-                vcpus.map { ceil((duration * 1000) / (it.vm.deadline - start) * it.burst).toLong() }.sum()
-            val totalRemainder = burst.sum()
-            val totalGrantedBurst = totalAllocatedBurst - totalRemainder
-
-            // The burst that was lost due to overcommissioning of CPU resources
-            var totalOvercommissionedBurst = 0L
-            // The burst that was lost due to interference.
-            var totalInterferedBurst = 0L
-
-            val vmIterator = vms.iterator()
-            while (vmIterator.hasNext()) {
-                val vm = vmIterator.next()
-
-                // Apply performance interference model
-                val performanceScore = vm.performanceInterferenceModel?.apply(serverLoad) ?: 1.0
-                var hasFinished = false
-
-                for (vcpu in vm.vcpus) {
-                    // Compute the fraction of compute time allocated to the VM
-                    val fraction = vcpu.allocatedLimit / totalAllocatedUsage
-
-                    // Compute the burst time that the VM was actually granted
-                    val grantedBurst = ceil(totalGrantedBurst * fraction).toLong()
-
-                    // The burst that was actually used by the VM
-                    val usedBurst = ceil(grantedBurst * performanceScore).toLong()
-
-                    totalInterferedBurst += grantedBurst - usedBurst
-
-                    // Compute remaining burst time to be executed for the request
-                    if (vcpu.consume(usedBurst)) {
-                        hasFinished = true
-                    } else if (vm.deadline <= end) {
-                        // Request must have its entire burst consumed or otherwise we have overcommission
-                        // Note that we count the overcommissioned burst if the hypervisor has failed.
-                        totalOvercommissionedBurst += vcpu.burst
-                    }
-                }
-
-                if (hasFinished || vm.deadline <= end) {
-                    // Mark the VM as finished and deschedule the VMs if needed
-                    if (vm.finish()) {
-                        vmIterator.remove()
-                        vcpus.removeAll(vm.vcpus)
-                    }
-                }
-            }
-
-            listener?.onSliceFinish(
-                this,
-                totalRequestedBurst,
-                min(totalRequestedSubBurst, totalGrantedBurst), // We can run more than requested due to timing
-                totalOvercommissionedBurst,
-                totalInterferedBurst, // Might be smaller than zero due to FP rounding errors,
-                min(
-                    totalAllocatedUsage,
-                    totalRequestedUsage
-                ), // The allocated usage might be slightly higher due to FP rounding
-                totalRequestedUsage
-            )
-        }
-    }
+    public fun createMachine(
+        model: SimMachineModel,
+        performanceInterferenceModel: PerformanceInterferenceModel? = null
+    ): SimMachine
 
     /**
      * Event listener for hypervisor events.
@@ -301,235 +57,4 @@ public class SimHypervisor(
             cpuDemand: Double
         )
     }
-
-    /**
-     * A scheduling command processed by the scheduler.
-     */
-    private sealed class SchedulerCommand {
-        /**
-         * Schedule the specified VM on the hypervisor.
-         */
-        data class Schedule(val vm: VmSession) : SchedulerCommand()
-
-        /**
-         * De-schedule the specified VM on the hypervisor.
-         */
-        data class Deschedule(val vm: VmSession) : SchedulerCommand()
-
-        /**
-         * Interrupt the scheduler.
-         */
-        object Interrupt : SchedulerCommand()
-    }
-
-    /**
-     * A virtual machine running on the hypervisor.
-     *
-     * @param ctx The execution context the vCPU runs in.
-     * @param triggerMode The mode when to trigger the VM exit.
-     * @param merge The function to merge consecutive slices on spillover.
-     * @param select The function to select on finish.
-     */
-    @OptIn(InternalCoroutinesApi::class)
-    private data class VmSession(
-        val model: SimMachineModel,
-        val performanceInterferenceModel: PerformanceInterferenceModel? = null,
-        var triggerMode: SimExecutionContext.TriggerMode = SimExecutionContext.TriggerMode.FIRST,
-        var merge: (SimExecutionContext.Slice, SimExecutionContext.Slice) -> SimExecutionContext.Slice = { _, r -> r },
-        var select: () -> Unit = {}
-    ) {
-        /**
-         * The vCPUs of this virtual machine.
-         */
-        val vcpus: List<VCpu>
-
-        /**
-         * The slices that the VM wants to run.
-         */
-        var queue: Iterator<SimExecutionContext.Slice> = emptyList<SimExecutionContext.Slice>().iterator()
-
-        /**
-         * The current active slice.
-         */
-        var activeSlice: SimExecutionContext.Slice? = null
-
-        /**
-         * The current deadline of the VM.
-         */
-        val deadline: Long
-            get() = activeSlice?.deadline ?: Long.MAX_VALUE
-
-        /**
-         * A flag to indicate that the VM is idle.
-         */
-        val isIdle: Boolean
-            get() = activeSlice == null
-
-        /**
-         * The usage of the virtual machine.
-         */
-        val usage: MutableStateFlow<Double> = MutableStateFlow(0.0)
-
-        init {
-            vcpus = model.cpus.mapIndexed { i, model -> VCpu(this, model, i) }
-        }
-
-        /**
-         * Schedule the given slices on this vCPU, replacing the existing slices.
-         */
-        fun schedule(slices: Sequence<SimExecutionContext.Slice>) {
-            queue = slices.iterator()
-
-            if (queue.hasNext()) {
-                activeSlice = queue.next()
-                refresh()
-            }
-        }
-
-        /**
-         * Cancel the existing workload on the VM.
-         */
-        fun cancel() {
-            queue = emptyList<SimExecutionContext.Slice>().iterator()
-            activeSlice = null
-            refresh()
-        }
-
-        /**
-         * Finish the current slice of the VM.
-         *
-         * @return `true` if the vCPUs may be descheduled, `false` otherwise.
-         */
-        fun finish(): Boolean {
-            val activeSlice = activeSlice ?: return true
-
-            return if (queue.hasNext()) {
-                val needsMerge = activeSlice.burst.any { it > 0 }
-                val candidateSlice = queue.next()
-                val slice = if (needsMerge) merge(activeSlice, candidateSlice) else candidateSlice
-
-                this.activeSlice = slice
-
-                // Update the vCPU cache
-                refresh()
-
-                false
-            } else {
-                this.activeSlice = null
-                select()
-                true
-            }
-        }
-
-        /**
-         * Refresh the vCPU cache.
-         */
-        fun refresh() {
-            vcpus.forEach { it.refresh() }
-            usage.value = vcpus.sumByDouble { it.burst / it.limit } / vcpus.size
-        }
-    }
-
-    /**
-     * A virtual CPU that can be scheduled on a physical CPU.
-     *
-     * @param vm The VM of which this vCPU is part.
-     * @param model The model of CPU that this vCPU models.
-     * @param id The id of the vCPU with respect to the VM.
-     */
-    private data class VCpu(
-        val vm: VmSession,
-        val model: ProcessingUnit,
-        val id: Int
-    ) : Comparable<VCpu> {
-        /**
-         * The current limit on the vCPU.
-         */
-        var limit: Double = 0.0
-
-        /**
-         * The limit allocated by the hypervisor.
-         */
-        var allocatedLimit: Double = 0.0
-
-        /**
-         * The current burst running on the vCPU.
-         */
-        var burst: Long = 0L
-
-        /**
-         * Consume the specified burst on this vCPU.
-         */
-        fun consume(burst: Long): Boolean {
-            this.burst = max(0, this.burst - burst)
-
-            // Flush the result to the slice if it exists
-            vm.activeSlice?.burst?.takeIf { id < it.size }?.set(id, this.burst)
-
-            return allocatedLimit > 0.0 && this.burst == 0L
-        }
-
-        /**
-         * Refresh the information of this vCPU based on the current slice.
-         */
-        fun refresh() {
-            limit = vm.activeSlice?.limit?.takeIf { id < it.size }?.get(id) ?: 0.0
-            burst = vm.activeSlice?.burst?.takeIf { id < it.size }?.get(id) ?: 0
-        }
-
-        /**
-         * Compare to another vCPU based on the current load of the vCPU.
-         */
-        override fun compareTo(other: VCpu): Int {
-            return limit.compareTo(other.limit)
-        }
-
-        /**
-         * Create a string representation of the vCPU.
-         */
-        override fun toString(): String =
-            "vCPU(id=$id,burst=$burst,limit=$limit,allocatedLimit=$allocatedLimit)"
-    }
-
-    /**
-     * The execution context in which a VM runs.
-     *
-     */
-    private inner class VmExecutionContext(val session: VmSession) :
-        SimExecutionContext, DisposableHandle {
-        override val machine: SimMachineModel
-            get() = session.model
-
-        override val clock: Clock
-            get() = this@SimHypervisor.clock
-
-        @OptIn(InternalCoroutinesApi::class)
-        override fun onRun(
-            batch: Sequence<SimExecutionContext.Slice>,
-            triggerMode: SimExecutionContext.TriggerMode,
-            merge: (SimExecutionContext.Slice, SimExecutionContext.Slice) -> SimExecutionContext.Slice
-        ): SelectClause0 = object : SelectClause0 {
-            @InternalCoroutinesApi
-            override fun <R> registerSelectClause0(select: SelectInstance<R>, block: suspend () -> R) {
-                session.triggerMode = triggerMode
-                session.merge = merge
-                session.select = {
-                    if (select.trySelect()) {
-                        block.startCoroutineCancellable(select.completion)
-                    }
-                }
-                session.schedule(batch)
-                // Indicate to the hypervisor that the VM should be re-scheduled
-                schedulingQueue.offer(SchedulerCommand.Schedule(session))
-                select.disposeOnSelect(this@VmExecutionContext)
-            }
-        }
-
-        override fun dispose() {
-            if (!session.isIdle) {
-                session.cancel()
-                schedulingQueue.offer(SchedulerCommand.Deschedule(session))
-            }
-        }
-    }
 }
diff --git a/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/workload/SimFlopsWorkload.kt b/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/workload/SimFlopsWorkload.kt
index 918a78bd..0d2c9374 100644
--- a/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/workload/SimFlopsWorkload.kt
+++ b/simulator/opendc-simulator/opendc-simulator-compute/src/main/kotlin/org/opendc/simulator/compute/workload/SimFlopsWorkload.kt
@@ -39,7 +39,7 @@ public class SimFlopsWorkload(
     public val utilization: Double = 0.8
 ) : SimWorkload {
     init {
-        require(flops >= 0) { "Negative number of flops" }
+        require(flops >= 0) { "Negative number of FLOPs" }
         require(cores > 0) { "Negative number of cores or no cores" }
         require(utilization > 0.0 && utilization <= 1.0) { "Utilization must be in (0, 1]" }
     }
@@ -54,4 +54,6 @@ public class SimFlopsWorkload(
 
         ctx.run(SimExecutionContext.Slice(burst, maxUsage, Long.MAX_VALUE), triggerMode = SimExecutionContext.TriggerMode.LAST)
     }
+
+    override fun toString(): String = "SimFlopsWorkload(FLOPs=$flops,cores=$cores,utilization=$utilization)"
 }