[ML] Remove out-of-phase buckets feature (#318)

This feature was never fully completed and in fact
we no longer need it as the multibucket feature
covers the benefits from supporting out-of-phase buckets.

Author: dimitris-athanasiou

bin/autodetect/CCmdLineParser.cc

@@ -50,7 +50,6 @@ bool CCmdLineParser::parse(int argc,
                            bool& isPersistFileNamedPipe,
                            size_t& maxAnomalyRecords,
                            bool& memoryUsage,
-                           std::size_t& bucketResultsDelay,
                            bool& multivariateByFields,
                            TStrVec& clauseTokens) {
     try {
@@ -111,8 +110,6 @@ bool CCmdLineParser::parse(int argc,
                         "The maximum number of records to be outputted for each bucket. Defaults to 100, a value 0 removes the limit.")
             ("memoryUsage",
                         "Log the model memory usage at the end of the job")
-            ("resultFinalizationWindow", boost::program_options::value<std::size_t>(),
-                        "The numer of half buckets to store before choosing which overlapping bucket has the biggest anomaly")
             ("multivariateByFields",
                         "Optional flag to enable multi-variate analysis of correlated by fields")
         ;
@@ -222,9 +219,6 @@ bool CCmdLineParser::parse(int argc,
         if (vm.count("memoryUsage") > 0) {
             memoryUsage = true;
         }
-        if (vm.count("resultFinalizationWindow") > 0) {
-            bucketResultsDelay = vm["resultFinalizationWindow"].as<std::size_t>();
-        }
         if (vm.count("multivariateByFields") > 0) {
             multivariateByFields = true;
         }

bin/autodetect/CCmdLineParser.h

@@ -62,7 +62,6 @@ class CCmdLineParser {
                       bool& isPersistFileNamedPipe,
                       size_t& maxAnomalyRecords,
                       bool& memoryUsage,
-                      std::size_t& bucketResultsDelay,
                       bool& multivariateByFields,
                       TStrVec& clauseTokens);
 

bin/autodetect/Main.cc

@@ -86,7 +86,6 @@ int main(int argc, char** argv) {
     bool isPersistFileNamedPipe(false);
     size_t maxAnomalyRecords(100u);
     bool memoryUsage(false);
-    std::size_t bucketResultsDelay(0);
     bool multivariateByFields(false);
     TStrVec clauseTokens;
     if (ml::autodetect::CCmdLineParser::parse(
@@ -96,8 +95,8 @@ int main(int argc, char** argv) {
             timeFormat, quantilesStateFile, deleteStateFiles, persistInterval,
             maxQuantileInterval, inputFileName, isInputFileNamedPipe, outputFileName,
             isOutputFileNamedPipe, restoreFileName, isRestoreFileNamedPipe,
-            persistFileName, isPersistFileNamedPipe, maxAnomalyRecords, memoryUsage,
-            bucketResultsDelay, multivariateByFields, clauseTokens) == false) {
+            persistFileName, isPersistFileNamedPipe, maxAnomalyRecords,
+            memoryUsage, multivariateByFields, clauseTokens) == false) {
         return EXIT_FAILURE;
     }
 
@@ -143,8 +142,7 @@ int main(int argc, char** argv) {
         summaryCountFieldName.empty() ? ml::model_t::E_None : ml::model_t::E_Manual);
     ml::model::CAnomalyDetectorModelConfig modelConfig =
         ml::model::CAnomalyDetectorModelConfig::defaultConfig(
-            bucketSpan, summaryMode, summaryCountFieldName, latency,
-            bucketResultsDelay, multivariateByFields);
+            bucketSpan, summaryMode, summaryCountFieldName, latency, multivariateByFields);
     modelConfig.detectionRules(ml::model::CAnomalyDetectorModelConfig::TIntDetectionRuleVecUMapCRef(
         fieldConfig.detectionRules()));
     modelConfig.scheduledEvents(ml::model::CAnomalyDetectorModelConfig::TStrDetectionRulePrVecCRef(

include/api/CAnomalyJob.h

@@ -12,13 +12,11 @@
 
 #include <model/CAnomalyDetector.h>
 #include <model/CAnomalyDetectorModelConfig.h>
-#include <model/CBucketQueue.h>
 #include <model/CHierarchicalResults.h>
 #include <model/CHierarchicalResultsAggregator.h>
 #include <model/CHierarchicalResultsNormalizer.h>
 #include <model/CInterimBucketCorrector.h>
 #include <model/CResourceMonitor.h>
-#include <model/CResultsQueue.h>
 #include <model/CSearchKey.h>
 
 #include <api/CDataProcessor.h>
@@ -117,26 +115,20 @@ class API_EXPORT CAnomalyJob : public CDataProcessor {
         std::pair<model::CSearchKey::TStrCRefKeyCRefPr, TAnomalyDetectorPtr>;
     using TKeyCRefAnomalyDetectorPtrPrVec = std::vector<TKeyCRefAnomalyDetectorPtrPr>;
     using TModelPlotDataVec = model::CAnomalyDetector::TModelPlotDataVec;
-    using TModelPlotDataVecCItr = TModelPlotDataVec::const_iterator;
-    using TModelPlotDataVecQueue = model::CBucketQueue<TModelPlotDataVec>;
 
     struct API_EXPORT SRestoredStateDetail {
         ERestoreStateStatus s_RestoredStateStatus;
         boost::optional<std::string> s_Extra;
     };
 
     struct SBackgroundPersistArgs {
-        SBackgroundPersistArgs(const model::CResultsQueue& resultsQueue,
-                               const TModelPlotDataVecQueue& modelPlotQueue,
-                               core_t::TTime time,
+        SBackgroundPersistArgs(core_t::TTime time,
                                const model::CResourceMonitor::SResults& modelSizeStats,
                                const model::CInterimBucketCorrector& interimBucketCorrector,
                                const model::CHierarchicalResultsAggregator& aggregator,
                                core_t::TTime latestRecordTime,
                                core_t::TTime lastResultsTime);
 
-        model::CResultsQueue s_ResultsQueue;
-        TModelPlotDataVecQueue s_ModelPlotQueue;
         core_t::TTime s_Time;
         model::CResourceMonitor::SResults s_ModelSizeStats;
         model::CInterimBucketCorrector s_InterimBucketCorrector;
@@ -224,14 +216,11 @@ class API_EXPORT CAnomalyJob : public CDataProcessor {
     void outputInterimResults(core_t::TTime bucketStartTime);
 
     //! Helper function for outputResults.
-    //! \p processingTimer is the processing time can be written to the bucket
-    //! \p sumPastProcessingTime is the total time previously spent processing
-    //! but resulted in no bucket being outputted.
+    //! \p processingTime is the processing time of the bucket
     void writeOutResults(bool interim,
                          model::CHierarchicalResults& results,
                          core_t::TTime bucketTime,
-                         uint64_t processingTime,
-                         uint64_t sumPastProcessingTime);
+                         uint64_t processingTime);
 
     //! Reset buckets in the range specified by the control message.
     void resetBuckets(const std::string& controlMessage);
@@ -259,8 +248,6 @@ class API_EXPORT CAnomalyJob : public CDataProcessor {
 
     //! Persist the detectors to a stream.
     bool persistState(const std::string& descriptionPrefix,
-                      const model::CResultsQueue& resultsQueue,
-                      const TModelPlotDataVecQueue& modelPlotQueue,
                       core_t::TTime time,
                       const TKeyCRefAnomalyDetectorPtrPrVec& detectors,
                       const model::CResourceMonitor::SResults& modelSizeStats,
@@ -296,16 +283,9 @@ class API_EXPORT CAnomalyJob : public CDataProcessor {
     //! \param[in] endTime The end of the time interval to skip sampling.
     void skipSampling(core_t::TTime endTime);
 
-    //! Outputs queued results and resets the queue to the given \p startTime
-    void flushAndResetResultsQueue(core_t::TTime startTime);
-
     //! Roll time forward to \p time
     void timeNow(core_t::TTime time);
 
-    //! Get the bucketLength, or half the bucketLength if
-    //! out-of-phase buckets are active
-    core_t::TTime effectiveBucketLength() const;
-
     //! Update configuration
     void updateConfig(const std::string& config);
 
@@ -333,15 +313,12 @@ class API_EXPORT CAnomalyJob : public CDataProcessor {
     //! specified time range.
     void generateModelPlot(core_t::TTime startTime,
                            core_t::TTime endTime,
-                           const model::CAnomalyDetector& detector);
+                           const model::CAnomalyDetector& detector,
+                           TModelPlotDataVec& modelPlotData);
 
     //! Write the pre-generated model plot to the output stream of the user's
     //! choosing: either file or streamed to the API
-    void writeOutModelPlot(core_t::TTime resultsTime);
-
-    //! Write the pre-generated model plot to the output stream of the user's
-    //! choosing: either file or streamed to the API
-    void writeOutModelPlot(core_t::TTime, CModelPlotDataJsonWriter& writer);
+    void writeOutModelPlot(const TModelPlotDataVec& modelPlotData);
 
     //! Persist one detector to a stream.
     //! This method is static so that there is no danger of it accessing
@@ -477,15 +454,6 @@ class API_EXPORT CAnomalyJob : public CDataProcessor {
     //! The hierarchical results normalizer.
     model::CHierarchicalResultsNormalizer m_Normalizer;
 
-    //! Store the last N half-buckets' results in order
-    //! to choose the best result
-    model::CResultsQueue m_ResultsQueue;
-
-    //! Also store the model plot for the buckets for each
-    //! result time - these will be output when the corresponding
-    //! result is output
-    TModelPlotDataVecQueue m_ModelPlotQueue;
-
     friend class ::CBackgroundPersisterTest;
     friend class ::CAnomalyJobTest;
 };

include/model/CAnomalyDetectorModel.h

@@ -352,17 +352,6 @@ class MODEL_EXPORT CAnomalyDetectorModel {
                         core_t::TTime endTime,
                         CResourceMonitor& resourceMonitor) = 0;
 
-    //! This samples the bucket statistics, and any state needed
-    //! by computeProbablity, in the time interval [\p startTime,
-    //! \p endTime], but does not update the model. This is needed
-    //! by the results preview.
-    //!
-    //! \param[in] startTime The start of the time interval to sample.
-    //! \param[in] endTime The end of the time interval to sample.
-    virtual void sampleOutOfPhase(core_t::TTime startTime,
-                                  core_t::TTime endTime,
-                                  CResourceMonitor& resourceMonitor) = 0;
-
     //! Rolls time to \p endTime while skipping sampling the models for
     //! buckets within the gap.
     //!

include/model/CAnomalyDetectorModelConfig.h

@@ -114,10 +114,6 @@ class MODEL_EXPORT CAnomalyDetectorModelConfig {
 
     //! Bucket length corresponding to the default decay and learn rates.
     static const core_t::TTime STANDARD_BUCKET_LENGTH;
-
-    //! The default number of half buckets to store before choosing which
-    //! overlapping bucket has the biggest anomaly
-    static const std::size_t DEFAULT_BUCKET_RESULTS_DELAY;
     //@}
 
     //! \name Modelling
@@ -239,15 +235,12 @@ class MODEL_EXPORT CAnomalyDetectorModelConfig {
     //! then this is the name of the field holding the summary count.
     //! \param[in] latency The amount of time records are buffered for, to
     //! allow out-of-order records to be seen by the models in order.
-    //! \param[in] bucketResultsDelay The number of half-bucket results
-    //! to sit on before giving a definitive result.
     //! \param[in] multivariateByFields Should multivariate analysis of
     //! correlated 'by' fields be performed?
     static CAnomalyDetectorModelConfig defaultConfig(core_t::TTime bucketLength,
                                                      model_t::ESummaryMode summaryMode,
                                                      const std::string& summaryCountFieldName,
                                                      core_t::TTime latency,
-                                                     std::size_t bucketResultsDelay,
                                                      bool multivariateByFields);
 
     //! Overload using defaults.
@@ -256,8 +249,7 @@ class MODEL_EXPORT CAnomalyDetectorModelConfig {
                   model_t::ESummaryMode summaryMode = model_t::E_None,
                   const std::string& summaryCountFieldName = "") {
         return defaultConfig(bucketLength, summaryMode, summaryCountFieldName,
-                             DEFAULT_LATENCY_BUCKETS * bucketLength,
-                             DEFAULT_BUCKET_RESULTS_DELAY, false);
+                             DEFAULT_LATENCY_BUCKETS * bucketLength, false);
     }
 
     //! Get the factor to normalize all bucket lengths to the default
@@ -273,8 +265,6 @@ class MODEL_EXPORT CAnomalyDetectorModelConfig {
 
     //! Set the data bucketing interval.
     void bucketLength(core_t::TTime length);
-    //! Set the number of buckets to delay finalizing out-of-phase buckets.
-    void bucketResultsDelay(std::size_t delay);
     //! Set the single interim bucket correction calculator.
     void interimBucketCorrector(const TInterimBucketCorrectorPtr& interimBucketCorrector);
     //! Set whether to model multibucket features.
@@ -358,9 +348,6 @@ class MODEL_EXPORT CAnomalyDetectorModelConfig {
     //! numbers of buckets.
     std::size_t latencyBuckets() const;
 
-    //! Get the bucket result delay window.
-    std::size_t bucketResultsDelay() const;
-
     //! Get the single interim bucket correction calculator.
     const CInterimBucketCorrector& interimBucketCorrector() const;
 
@@ -444,10 +431,6 @@ class MODEL_EXPORT CAnomalyDetectorModelConfig {
     //! Bucket length.
     core_t::TTime m_BucketLength;
 
-    //! Get the bucket result delay window: The numer of half buckets to
-    //! store before choosing which overlapping bucket has the biggest anomaly
-    std::size_t m_BucketResultsDelay;
-
     //! Should multivariate analysis of correlated 'by' fields be performed?
     bool m_MultivariateByFields;
 

include/model/CBucketGatherer.h

@@ -56,7 +56,8 @@ class CResourceMonitor;
 //! IMPLEMENTATION:\n
 //! This functionality has been separated from the CDataGatherer in order
 //! to allow the CDataGatherer to support multiple overlapping buckets and
-//! buckets with different time spans.
+//! buckets with different time spans. However, the overlapping feature
+//! has been removed but this class is kept to avoid BWC issues.
 class MODEL_EXPORT CBucketGatherer {
 public:
     using TDoubleVec = std::vector<double>;
@@ -394,6 +395,9 @@ class MODEL_EXPORT CBucketGatherer {
     //! Create samples if possible for the bucket pointed out by \p time.
     virtual void sample(core_t::TTime time) = 0;
 
+    //! Persist state by passing information \p inserter.
+    virtual void acceptPersistInserter(core::CStatePersistInserter& inserter) const = 0;
+
 private:
     //! Resize the necessary data structures so they can hold values
     //! for the person and/or attribute identified by \p pid and \p cid,

include/model/CCountingModel.h

@@ -160,17 +160,6 @@ class MODEL_EXPORT CCountingModel : public CAnomalyDetectorModel {
                                         core_t::TTime endTime,
                                         CResourceMonitor& resourceMonitor);
 
-    //! This samples the bucket statistics, and any state needed
-    //! by computeProbablity, in the time interval [\p startTime,
-    //! \p endTime], but does not update the model. This is needed
-    //! by the results preview.
-    //!
-    //! \param[in] startTime The start of the time interval to sample.
-    //! \param[in] endTime The end of the time interval to sample.
-    virtual void sampleOutOfPhase(core_t::TTime startTime,
-                                  core_t::TTime endTime,
-                                  CResourceMonitor& resourceMonitor);
-
     //! This samples the bucket statistics, in the time interval
     //! [\p startTime, \p endTime].
     //!

include/model/CCountingModelFactory.h

@@ -117,9 +117,6 @@ class MODEL_EXPORT CCountingModelFactory : public CModelFactory {
 
     //! Set the features which will be modeled.
     virtual void features(const TFeatureVec& features);
-
-    //! Set the bucket results delay
-    virtual void bucketResultsDelay(std::size_t bucketResultsDelay);
     //@}
 
     //! Get the minimum seasonal variance scale
@@ -153,9 +150,6 @@ class MODEL_EXPORT CCountingModelFactory : public CModelFactory {
     //! The count features which will be modeled.
     TFeatureVec m_Features;
 
-    //! The bucket results delay.
-    std::size_t m_BucketResultsDelay;
-
     //! A cached search key.
     mutable TOptionalSearchKey m_SearchKeyCache;
 };

include/model/CDataGatherer.h

@@ -120,7 +120,6 @@ class MODEL_EXPORT CDataGatherer {
         CBucketQueue<TSizeSizePrStoredStringPtrPrUInt64UMapVec>;
     using TSearchKeyCRef = boost::reference_wrapper<const CSearchKey>;
     using TBucketGathererPtr = std::unique_ptr<CBucketGatherer>;
-    using TBucketGathererPtrVec = std::vector<TBucketGathererPtr>;
     using TFeatureAnyPr = std::pair<model_t::EFeature, boost::any>;
     using TFeatureAnyPrVec = std::vector<TFeatureAnyPr>;
     using TMetricCategoryVec = std::vector<model_t::EMetricCategory>;
@@ -338,7 +337,7 @@ class MODEL_EXPORT CDataGatherer {
                      core_t::TTime bucketLength,
                      std::vector<std::pair<model_t::EFeature, T>>& result) const {
         TFeatureAnyPrVec rawFeatureData;
-        this->chooseBucketGatherer(time).featureData(time, bucketLength, rawFeatureData);
+        m_BucketGatherer->featureData(time, bucketLength, rawFeatureData);
 
         bool succeeded = true;
 
@@ -558,7 +557,7 @@ class MODEL_EXPORT CDataGatherer {
     void timeNow(core_t::TTime time);
 
     //! Print the current bucket.
-    std::string printCurrentBucket(core_t::TTime time) const;
+    std::string printCurrentBucket() const;
 
     //! Record a attribute called \p attribute.
     std::size_t addAttribute(const std::string& attribute,
@@ -591,9 +590,6 @@ class MODEL_EXPORT CDataGatherer {
 
     //! Reset bucket and return true if bucket was successfully
     //! reset or false otherwise.
-    //! Note that this should not be used in conjunction with out-of-phase buckets
-    //! where the concept of resetting a specific bucketed period of time is
-    //! not valid.
     bool resetBucket(core_t::TTime bucketStart);
 
     //! Release memory that is no longer needed
@@ -693,14 +689,6 @@ class MODEL_EXPORT CDataGatherer {
     using TModelParamsCRef = boost::reference_wrapper<const SModelParams>;
 
 private:
-    //! Select the correct bucket gatherer based on the time: if we have
-    //! out-of-phase buckets, select either in-phase or out-of-phase.
-    const CBucketGatherer& chooseBucketGatherer(core_t::TTime time) const;
-
-    //! Select the correct bucket gatherer based on the time: if we have
-    //! out-of-phase buckets, select either in-phase or out-of-phase.
-    CBucketGatherer& chooseBucketGatherer(core_t::TTime time);
-
     //! Restore state from supplied traverser.
     bool acceptRestoreTraverser(const std::string& summaryCountFieldName,
                                 const std::string& personFieldName,
@@ -738,9 +726,9 @@ class MODEL_EXPORT CDataGatherer {
     //! The collection of features on which to gather data.
     TFeatureVec m_Features;
 
-    //! The collection of bucket gatherers which contain the bucket-specific
+    //! The bucket gatherer which contains the bucket-specific
     //! metrics and counts.
-    TBucketGathererPtrVec m_Gatherers;
+    TBucketGathererPtr m_BucketGatherer;
 
     //! Indicates whether the data being gathered are already summarized
     //! by an external aggregation process.

include/model/CEventRateBucketGatherer.h

@@ -155,7 +155,7 @@ class MODEL_EXPORT CEventRateBucketGatherer final : public CBucketGatherer {
     bool acceptRestoreTraverser(core::CStateRestoreTraverser& traverser);
 
     //! Persist state by passing information to the supplied inserter
-    void acceptPersistInserter(core::CStatePersistInserter& inserter) const;
+    virtual void acceptPersistInserter(core::CStatePersistInserter& inserter) const;
 
     //! Create a clone of this data gatherer that will result in the same
     //! persisted state.  The clone may be incomplete in ways that do not