// Licensed to the .NET Foundation under one or more agreements. // The .NET Foundation licenses this file to you under the MIT license. // See the LICENSE file in the project root for more information. //========================================================================= // // HillClimbing.cpp // // Defines classes for the ThreadPool's HillClimbing concurrency-optimization // algorithm. // //========================================================================= // // TODO: write an essay about how/why this works. Maybe put it in BotR? // #include "common.h" #include "hillclimbing.h" #include "win32threadpool.h" // // Default compilation mode is /fp:precise, which disables fp intrinsics. This causes us to pull in FP stuff (sin,cos,etc.) from // The CRT, and increases our download size by ~5k. We don't need the extra precision this gets us, so let's switch to // the intrinsic versions. // #ifdef _MSC_VER #pragma float_control(precise, off) #endif const double pi = 3.141592653589793; void HillClimbing::Initialize() { CONTRACTL { THROWS; GC_NOTRIGGER; MODE_ANY; } CONTRACTL_END; m_wavePeriod = CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_WavePeriod); m_maxThreadWaveMagnitude = CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_MaxWaveMagnitude); m_threadMagnitudeMultiplier = (double)CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_WaveMagnitudeMultiplier) / 100.0; m_samplesToMeasure = m_wavePeriod * (int)CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_WaveHistorySize); m_targetThroughputRatio = (double)CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_Bias) / 100.0; m_targetSignalToNoiseRatio = (double)CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_TargetSignalToNoiseRatio) / 100.0; m_maxChangePerSecond = (double)CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_MaxChangePerSecond); m_maxChangePerSample = (double)CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_MaxChangePerSample); m_sampleIntervalLow = CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_SampleIntervalLow); m_sampleIntervalHigh = CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_SampleIntervalHigh); m_throughputErrorSmoothingFactor = (double)CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_ErrorSmoothingFactor) / 100.0; m_gainExponent = (double)CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_GainExponent) / 100.0; m_maxSampleError = (double)CLRConfig::GetConfigValue(CLRConfig::INTERNAL_HillClimbing_MaxSampleErrorPercent) / 100.0; m_currentControlSetting = 0; m_totalSamples = 0; m_lastThreadCount = 0; m_averageThroughputNoise = 0; m_elapsedSinceLastChange = 0; m_completionsSinceLastChange = 0; m_accumulatedCompletionCount = 0; m_accumulatedSampleDuration = 0; m_samples = new double[m_samplesToMeasure]; m_threadCounts = new double[m_samplesToMeasure]; // seed our random number generator with the CLR instance ID and the process ID, to avoid correlations with other CLR ThreadPool instances. #ifndef DACCESS_COMPILE m_randomIntervalGenerator.Init(((int)GetClrInstanceId() << 16) ^ (int)GetCurrentProcessId()); #endif m_currentSampleInterval = m_randomIntervalGenerator.Next(m_sampleIntervalLow, m_sampleIntervalHigh+1); } int HillClimbing::Update(int currentThreadCount, double sampleDuration, int numCompletions, int* pNewSampleInterval) { LIMITED_METHOD_CONTRACT; #ifdef DACCESS_COMPILE return 1; #else // // If someone changed the thread count without telling us, update our records accordingly. // if (currentThreadCount != m_lastThreadCount) ForceChange(currentThreadCount, Initializing); // // Update the cumulative stats for this thread count // m_elapsedSinceLastChange += sampleDuration; m_completionsSinceLastChange += numCompletions; // // Add in any data we've already collected about this sample // sampleDuration += m_accumulatedSampleDuration; numCompletions += m_accumulatedCompletionCount; // // We need to make sure we're collecting reasonably accurate data. Since we're just counting the end // of each work item, we are goinng to be missing some data about what really happened during the // sample interval. The count produced by each thread includes an initial work item that may have // started well before the start of the interval, and each thread may have been running some new // work item for some time before the end of the interval, which did not yet get counted. So // our count is going to be off by +/- threadCount workitems. // // The exception is that the thread that reported to us last time definitely wasn't running any work // at that time, and the thread that's reporting now definitely isn't running a work item now. So // we really only need to consider threadCount-1 threads. // // Thus the percent error in our count is +/- (threadCount-1)/numCompletions. // // We cannot rely on the frequency-domain analysis we'll be doing later to filter out this error, because // of the way it accumulates over time. If this sample is off by, say, 33% in the negative direction, // then the next one likely will be too. The one after that will include the sum of the completions // we missed in the previous samples, and so will be 33% positive. So every three samples we'll have // two "low" samples and one "high" sample. This will appear as periodic variation right in the frequency // range we're targeting, which will not be filtered by the frequency-domain translation. // if (m_totalSamples > 0 && ((currentThreadCount-1.0) / numCompletions) >= m_maxSampleError) { // not accurate enough yet. Let's accumulate the data so far, and tell the ThreadPool // to collect a little more. m_accumulatedSampleDuration = sampleDuration; m_accumulatedCompletionCount = numCompletions; *pNewSampleInterval = 10; return currentThreadCount; } // // We've got enouugh data for our sample; reset our accumulators for next time. // m_accumulatedSampleDuration = 0; m_accumulatedCompletionCount = 0; // // Add the current thread count and throughput sample to our history // double throughput = (double)numCompletions / sampleDuration; FireEtwThreadPoolWorkerThreadAdjustmentSample(throughput, GetClrInstanceId()); int sampleIndex = m_totalSamples % m_samplesToMeasure; m_samples[sampleIndex] = throughput; m_threadCounts[sampleIndex] = currentThreadCount; m_totalSamples++; // // Set up defaults for our metrics // Complex threadWaveComponent = 0; Complex throughputWaveComponent = 0; double throughputErrorEstimate = 0; Complex ratio = 0; double confidence = 0; HillClimbingStateTransition transition = Warmup; // // How many samples will we use? It must be at least the three wave periods we're looking for, and it must also be a whole // multiple of the primary wave's period; otherwise the frequency we're looking for will fall between two frequency bands // in the Fourier analysis, and we won't be able to measure it accurately. // int sampleCount = ((int)min(m_totalSamples-1, m_samplesToMeasure) / m_wavePeriod) * m_wavePeriod; if (sampleCount > m_wavePeriod) { // // Average the throughput and thread count samples, so we can scale the wave magnitudes later. // double sampleSum = 0; double threadSum = 0; for (int i = 0; i < sampleCount; i++) { sampleSum += m_samples[(m_totalSamples - sampleCount + i) % m_samplesToMeasure]; threadSum += m_threadCounts[(m_totalSamples - sampleCount + i) % m_samplesToMeasure]; } double averageThroughput = sampleSum / sampleCount; double averageThreadCount = threadSum / sampleCount; if (averageThroughput > 0 && averageThreadCount > 0) { // // Calculate the periods of the adjacent frequency bands we'll be using to measure noise levels. // We want the two adjacent Fourier frequency bands. // double adjacentPeriod1 = sampleCount / (((double)sampleCount / (double)m_wavePeriod) + 1); double adjacentPeriod2 = sampleCount / (((double)sampleCount / (double)m_wavePeriod) - 1); // // Get the the three different frequency components of the throughput (scaled by average // throughput). Our "error" estimate (the amount of noise that might be present in the // frequency band we're really interested in) is the average of the adjacent bands. // throughputWaveComponent = GetWaveComponent(m_samples, sampleCount, m_wavePeriod) / averageThroughput; throughputErrorEstimate = abs(GetWaveComponent(m_samples, sampleCount, adjacentPeriod1) / averageThroughput); if (adjacentPeriod2 <= sampleCount) throughputErrorEstimate = max(throughputErrorEstimate, abs(GetWaveComponent(m_samples, sampleCount, adjacentPeriod2) / averageThroughput)); // // Do the same for the thread counts, so we have something to compare to. We don't measure thread count // noise, because there is none; these are exact measurements. // threadWaveComponent = GetWaveComponent(m_threadCounts, sampleCount, m_wavePeriod) / averageThreadCount; // // Update our moving average of the throughput noise. We'll use this later as feedback to // determine the new size of the thread wave. // if (m_averageThroughputNoise == 0) m_averageThroughputNoise = throughputErrorEstimate; else m_averageThroughputNoise = (m_throughputErrorSmoothingFactor * throughputErrorEstimate) + ((1.0-m_throughputErrorSmoothingFactor) * m_averageThroughputNoise); if (abs(threadWaveComponent) > 0) { // // Adjust the throughput wave so it's centered around the target wave, and then calculate the adjusted throughput/thread ratio. // ratio = (throughputWaveComponent - (m_targetThroughputRatio * threadWaveComponent)) / threadWaveComponent; transition = ClimbingMove; } else { ratio = 0; transition = Stabilizing; } // // Calculate how confident we are in the ratio. More noise == less confident. This has // the effect of slowing down movements that might be affected by random noise. // double noiseForConfidence = max(m_averageThroughputNoise, throughputErrorEstimate); if (noiseForConfidence > 0) confidence = (abs(threadWaveComponent) / noiseForConfidence) / m_targetSignalToNoiseRatio; else confidence = 1.0; //there is no noise! } } // // We use just the real part of the complex ratio we just calculated. If the throughput signal // is exactly in phase with the thread signal, this will be the same as taking the magnitude of // the complex move and moving that far up. If they're 180 degrees out of phase, we'll move // backward (because this indicates that our changes are having the opposite of the intended effect). // If they're 90 degrees out of phase, we won't move at all, because we can't tell wether we're // having a negative or positive effect on throughput. // double move = min(1.0, max(-1.0, ratio.r)); // // Apply our confidence multiplier. // move *= min(1.0, max(0.0, confidence)); // // Now apply non-linear gain, such that values around zero are attenuated, while higher values // are enhanced. This allows us to move quickly if we're far away from the target, but more slowly // if we're getting close, giving us rapid ramp-up without wild oscillations around the target. // double gain = m_maxChangePerSecond * sampleDuration; move = pow(fabs(move), m_gainExponent) * (move >= 0.0 ? 1 : -1) * gain; move = min(move, m_maxChangePerSample); // // If the result was positive, and CPU is > 95%, refuse the move. // if (move > 0.0 && ThreadpoolMgr::cpuUtilization > CpuUtilizationHigh) move = 0.0; // // Apply the move to our control setting // m_currentControlSetting += move; // // Calculate the new thread wave magnitude, which is based on the moving average we've been keeping of // the throughput error. This average starts at zero, so we'll start with a nice safe little wave at first. // int newThreadWaveMagnitude = (int)(0.5 + (m_currentControlSetting * m_averageThroughputNoise * m_targetSignalToNoiseRatio * m_threadMagnitudeMultiplier * 2.0)); newThreadWaveMagnitude = min(newThreadWaveMagnitude, m_maxThreadWaveMagnitude); newThreadWaveMagnitude = max(newThreadWaveMagnitude, 1); // // Make sure our control setting is within the ThreadPool's limits // m_currentControlSetting = min(ThreadpoolMgr::MaxLimitTotalWorkerThreads-newThreadWaveMagnitude, m_currentControlSetting); m_currentControlSetting = max(ThreadpoolMgr::MinLimitTotalWorkerThreads, m_currentControlSetting); // // Calculate the new thread count (control setting + square wave) // int newThreadCount = (int)(m_currentControlSetting + newThreadWaveMagnitude * ((m_totalSamples / (m_wavePeriod/2)) % 2)); // // Make sure the new thread count doesn't exceed the ThreadPool's limits // newThreadCount = min(ThreadpoolMgr::MaxLimitTotalWorkerThreads, newThreadCount); newThreadCount = max(ThreadpoolMgr::MinLimitTotalWorkerThreads, newThreadCount); // // Record these numbers for posterity // FireEtwThreadPoolWorkerThreadAdjustmentStats( sampleDuration, throughput, threadWaveComponent.r, throughputWaveComponent.r, throughputErrorEstimate, m_averageThroughputNoise, ratio.r, confidence, m_currentControlSetting, (unsigned short)newThreadWaveMagnitude, GetClrInstanceId()); // // If all of this caused an actual change in thread count, log that as well. // if (newThreadCount != currentThreadCount) ChangeThreadCount(newThreadCount, transition); // // Return the new thread count and sample interval. This is randomized to prevent correlations with other periodic // changes in throughput. Among other things, this prevents us from getting confused by Hill Climbing instances // running in other processes. // // If we're at minThreads, and we seem to be hurting performance by going higher, we can't go any lower to fix this. So // we'll simply stay at minThreads much longer, and only occasionally try a higher value. // if (ratio.r < 0.0 && newThreadCount == ThreadpoolMgr::MinLimitTotalWorkerThreads) *pNewSampleInterval = (int)(0.5 + m_currentSampleInterval * (10.0 * min(-ratio.r, 1.0))); else *pNewSampleInterval = m_currentSampleInterval; return newThreadCount; #endif //DACCESS_COMPILE } void HillClimbing::ForceChange(int newThreadCount, HillClimbingStateTransition transition) { LIMITED_METHOD_CONTRACT; if (newThreadCount != m_lastThreadCount) { m_currentControlSetting += (newThreadCount - m_lastThreadCount); ChangeThreadCount(newThreadCount, transition); } } void HillClimbing::ChangeThreadCount(int newThreadCount, HillClimbingStateTransition transition) { LIMITED_METHOD_CONTRACT; m_lastThreadCount = newThreadCount; m_currentSampleInterval = m_randomIntervalGenerator.Next(m_sampleIntervalLow, m_sampleIntervalHigh+1); double throughput = (m_elapsedSinceLastChange > 0) ? (m_completionsSinceLastChange / m_elapsedSinceLastChange) : 0; LogTransition(newThreadCount, throughput, transition); m_elapsedSinceLastChange = 0; m_completionsSinceLastChange = 0; } GARY_IMPL(HillClimbingLogEntry, HillClimbingLog, HillClimbingLogCapacity); GVAL_IMPL(int, HillClimbingLogFirstIndex); GVAL_IMPL(int, HillClimbingLogSize); void HillClimbing::LogTransition(int threadCount, double throughput, HillClimbingStateTransition transition) { LIMITED_METHOD_CONTRACT; #ifndef DACCESS_COMPILE int index = (HillClimbingLogFirstIndex + HillClimbingLogSize) % HillClimbingLogCapacity; if (HillClimbingLogSize == HillClimbingLogCapacity) { HillClimbingLogFirstIndex = (HillClimbingLogFirstIndex + 1) % HillClimbingLogCapacity; HillClimbingLogSize--; //hide this slot while we update it } HillClimbingLogEntry* entry = &HillClimbingLog[index]; entry->TickCount = GetTickCount(); entry->Transition = transition; entry->NewControlSetting = threadCount; entry->LastHistoryCount = (int)(min(m_totalSamples, m_samplesToMeasure) / m_wavePeriod) * m_wavePeriod; entry->LastHistoryMean = (float) throughput; HillClimbingLogSize++; FireEtwThreadPoolWorkerThreadAdjustmentAdjustment( throughput, threadCount, transition, GetClrInstanceId()); #endif //DACCESS_COMPILE } Complex HillClimbing::GetWaveComponent(double* samples, int sampleCount, double period) { LIMITED_METHOD_CONTRACT; _ASSERTE(sampleCount >= period); //can't measure a wave that doesn't fit _ASSERTE(period >= 2); //can't measure above the Nyquist frequency // // Calculate the sinusoid with the given period. // We're using the Goertzel algorithm for this. See http://en.wikipedia.org/wiki/Goertzel_algorithm. // double w = 2.0 * pi / period; double cosine = cos(w); double sine = sin(w); double coeff = 2.0 * cosine; double q0 = 0, q1 = 0, q2 = 0; for (int i = 0; i < sampleCount; i++) { double sample = samples[(m_totalSamples - sampleCount + i) % m_samplesToMeasure]; q0 = coeff * q1 - q2 + sample; q2 = q1; q1 = q0; } return Complex(q1 - q2 * cosine, q2 * sine) / (double)sampleCount; }