3.0.0 Breaking Changes For KNN

Signed-off-by: Vikasht34 <viktari@amazon.com>

CHANGELOG.md

@@ -17,6 +17,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 ### Refactoring
 * Small Refactor Post Lucene 10.0.1 upgrade [#2541](https://github.com/opensearch-project/k-NN/pull/2541)
 * Refactor codec to leverage backwards_codecs [#2546](https://github.com/opensearch-project/k-NN/pull/2546)
+* 3.0.0 Breaking Changes For KNN [#2564] (https://github.com/opensearch-project/k-NN/pull/2564)
 
 ## [Unreleased 2.x](https://github.com/opensearch-project/k-NN/compare/2.19...2.x)
 ### Features

src/main/java/org/opensearch/knn/index/KNNSettings.java

@@ -68,10 +68,7 @@ public class KNNSettings {
     /**
      * Settings name
      */
-    public static final String KNN_SPACE_TYPE = "index.knn.space_type";
     public static final String INDEX_KNN_ADVANCED_APPROXIMATE_THRESHOLD = "index.knn.advanced.approximate_threshold";
-    public static final String KNN_ALGO_PARAM_M = "index.knn.algo_param.m";
-    public static final String KNN_ALGO_PARAM_EF_CONSTRUCTION = "index.knn.algo_param.ef_construction";
     public static final String KNN_ALGO_PARAM_EF_SEARCH = "index.knn.algo_param.ef_search";
     public static final String KNN_ALGO_PARAM_INDEX_THREAD_QTY = "knn.algo_param.index_thread_qty";
     public static final String KNN_MEMORY_CIRCUIT_BREAKER_ENABLED = "knn.memory.circuit_breaker.enabled";
@@ -80,7 +77,6 @@ public class KNNSettings {
     public static final String KNN_CIRCUIT_BREAKER_TRIGGERED = "knn.circuit_breaker.triggered";
     public static final String KNN_CACHE_ITEM_EXPIRY_ENABLED = "knn.cache.item.expiry.enabled";
     public static final String KNN_CACHE_ITEM_EXPIRY_TIME_MINUTES = "knn.cache.item.expiry.minutes";
-    public static final String KNN_PLUGIN_ENABLED = "knn.plugin.enabled";
     public static final String KNN_CIRCUIT_BREAKER_UNSET_PERCENTAGE = "knn.circuit_breaker.unset.percentage";
     public static final String KNN_INDEX = "index.knn";
     public static final String MODEL_INDEX_NUMBER_OF_SHARDS = "knn.model.index.number_of_shards";
@@ -159,14 +155,6 @@ public class KNNSettings {
         Setting.Property.NodeScope
     );
 
-    public static final Setting<String> INDEX_KNN_SPACE_TYPE = Setting.simpleString(
-        KNN_SPACE_TYPE,
-        INDEX_KNN_DEFAULT_SPACE_TYPE,
-        new SpaceTypeValidator(),
-        IndexScope,
-        Setting.Property.Deprecated
-    );
-
     /**
      * build_vector_data_structure_threshold - This parameter determines when to build vector data structure for knn fields during indexing
      * and merging. Setting -1 (min) will skip building graph, whereas on any other values, the graph will be built if
@@ -182,20 +170,6 @@ public class KNNSettings {
         Dynamic
     );
 
-    /**
-     * M - the number of bi-directional links created for every new element during construction.
-     * Reasonable range for M is 2-100. Higher M work better on datasets with high intrinsic
-     * dimensionality and/or high recall, while low M work better for datasets with low intrinsic dimensionality and/or low recalls.
-     * The parameter also determines the algorithm's memory consumption, which is roughly M * 8-10 bytes per stored element.
-     */
-    public static final Setting<Integer> INDEX_KNN_ALGO_PARAM_M_SETTING = Setting.intSetting(
-        KNN_ALGO_PARAM_M,
-        INDEX_KNN_DEFAULT_ALGO_PARAM_M,
-        2,
-        IndexScope,
-        Setting.Property.Deprecated
-    );
-
     /**
      *  ef or efSearch - the size of the dynamic list for the nearest neighbors (used during the search).
      *  Higher ef leads to more accurate but slower search. ef cannot be set lower than the number of queried nearest neighbors k.
@@ -209,18 +183,6 @@ public class KNNSettings {
         Dynamic
     );
 
-    /**
-     * ef_constrution - the parameter has the same meaning as ef, but controls the index_time/index_accuracy.
-     * Bigger ef_construction leads to longer construction(more indexing time), but better index quality.
-     */
-    public static final Setting<Integer> INDEX_KNN_ALGO_PARAM_EF_CONSTRUCTION_SETTING = Setting.intSetting(
-        KNN_ALGO_PARAM_EF_CONSTRUCTION,
-        INDEX_KNN_DEFAULT_ALGO_PARAM_EF_CONSTRUCTION,
-        2,
-        IndexScope,
-        Setting.Property.Deprecated
-    );
-
     public static final Setting<Integer> MODEL_INDEX_NUMBER_OF_SHARDS_SETTING = Setting.intSetting(
         MODEL_INDEX_NUMBER_OF_SHARDS,
         1,
@@ -393,11 +355,6 @@ public class KNNSettings {
      */
     public static Map<String, Setting<?>> dynamicCacheSettings = new HashMap<String, Setting<?>>() {
         {
-            /**
-             * KNN plugin enable/disable setting
-             */
-            put(KNN_PLUGIN_ENABLED, Setting.boolSetting(KNN_PLUGIN_ENABLED, true, NodeScope, Dynamic));
-
             /**
              * Weight circuit breaker settings
              */
@@ -555,10 +512,7 @@ private Setting<?> getSetting(String key) {
 
     public List<Setting<?>> getSettings() {
         List<Setting<?>> settings = Arrays.asList(
-            INDEX_KNN_SPACE_TYPE,
             INDEX_KNN_ADVANCED_APPROXIMATE_THRESHOLD_SETTING,
-            INDEX_KNN_ALGO_PARAM_M_SETTING,
-            INDEX_KNN_ALGO_PARAM_EF_CONSTRUCTION_SETTING,
             INDEX_KNN_ALGO_PARAM_EF_SEARCH_SETTING,
             KNN_ALGO_PARAM_INDEX_THREAD_QTY_SETTING,
             KNN_CIRCUIT_BREAKER_TRIGGERED_SETTING,
@@ -583,10 +537,6 @@ public List<Setting<?>> getSettings() {
             .collect(Collectors.toList());
     }
 
-    public static boolean isKNNPluginEnabled() {
-        return KNNSettings.state().getSettingValue(KNNSettings.KNN_PLUGIN_ENABLED);
-    }
-
     public static boolean isCircuitBreakerTriggered() {
         return KNNSettings.state().getSettingValue(KNNSettings.KNN_CIRCUIT_BREAKER_TRIGGERED);
     }

src/main/java/org/opensearch/knn/index/engine/SpaceTypeResolver.java

@@ -8,7 +8,6 @@
 import org.apache.logging.log4j.util.Strings;
 import org.opensearch.common.settings.Settings;
 import org.opensearch.index.mapper.MapperParsingException;
-import org.opensearch.knn.index.KNNSettings;
 import org.opensearch.knn.index.SpaceType;
 import org.opensearch.knn.index.VectorDataType;
 
@@ -49,12 +48,6 @@ public SpaceType resolveSpaceType(
         // 1. We try to get it from index setting, which is a relic of legacy.
         // 2. Otherwise, we return a default one.
         if (isSpaceTypeConfigured(methodSpaceType) == false && isSpaceTypeConfigured(topLevelSpaceType) == false) {
-            if (indexSettings != null) {
-                final String spaceType = indexSettings.get(KNNSettings.INDEX_KNN_SPACE_TYPE.getKey());
-                if (spaceType != null) {
-                    return SpaceType.getSpace(spaceType);
-                }
-            }
             return getDefaultSpaceType(vectorDataType);
         }
 

src/main/java/org/opensearch/knn/index/mapper/KNNVectorFieldMapper.java

@@ -60,7 +60,6 @@
 import static org.opensearch.knn.index.mapper.KNNVectorFieldMapperUtil.createStoredFieldForFloatVector;
 import static org.opensearch.knn.index.mapper.KNNVectorFieldMapperUtil.useFullFieldNameValidation;
 import static org.opensearch.knn.index.mapper.KNNVectorFieldMapperUtil.validateIfCircuitBreakerIsNotTriggered;
-import static org.opensearch.knn.index.mapper.KNNVectorFieldMapperUtil.validateIfKNNPluginEnabled;
 import static org.opensearch.knn.index.mapper.ModelFieldMapper.UNSET_MODEL_DIMENSION_IDENTIFIER;
 
 /**
@@ -425,6 +424,8 @@ public Mapper.Builder<?> parse(String name, Map<String, Object> node, ParserCont
                 // Resolve method component. For the legacy case where space type can be configured at index level,
                 // it first tries to use the given one then tries to get it from index setting when the space type is UNDEFINED.
                 resolveKNNMethodComponents(builder, parserContext, resolvedSpaceType);
+                // Validate if the KNN engine is allowed for index creation
+                validateBlockedKNNEngine(builder.knnMethodContext.get(), parserContext.indexVersionCreated());
                 validateFromKNNMethod(builder);
             }
 
@@ -457,6 +458,25 @@ private void validateModeAndCompressionForDataType(KNNVectorFieldMapper.Builder
             }
         }
 
+        /**
+         * Validates whether the provided KNN engine is allowed for index creation.
+         * If the engine is NMSLIB and the OpenSearch version is 3.0.0 or later,
+         * it throws an IllegalArgumentException to prevent new index creation.
+         *
+         * @param knnMethodContext The KNN method configuration that contains the engine type.
+         * @param indexVersionCreated The OpenSearch version when the index is being created.
+         * @throws IllegalArgumentException if the engine is NMSLIB and version is 3.0.0 or later.
+         */
+        private void validateBlockedKNNEngine(KNNMethodContext knnMethodContext, Version indexVersionCreated) {
+            if (knnMethodContext != null
+                && knnMethodContext.getKnnEngine() == KNNEngine.NMSLIB
+                && indexVersionCreated.onOrAfter(Version.V_3_0_0)) {
+                throw new IllegalArgumentException(
+                    "NMSLIB engine is deprecated and cannot be used for new index creation in OpenSearch 3.0.0 and later."
+                );
+            }
+        }
+
         private void validateFromFlat(KNNVectorFieldMapper.Builder builder) {
             if (builder.modelId.get() != null || builder.knnMethodContext.get() != null) {
                 throw new IllegalArgumentException("Cannot set modelId or method parameters when index.knn setting is false");
@@ -686,7 +706,6 @@ protected List<Field> getFieldsForByteVector(final byte[] array) {
      * Validation checks before parsing of doc begins
      */
     protected void validatePreparse() {
-        validateIfKNNPluginEnabled();
         validateIfCircuitBreakerIsNotTriggered();
     }
 

src/main/java/org/opensearch/knn/index/mapper/KNNVectorFieldMapperUtil.java

@@ -33,13 +33,10 @@
 import java.util.Arrays;
 import java.util.Map;
 
-import static org.opensearch.knn.common.KNNConstants.HNSW_ALGO_EF_CONSTRUCTION;
-import static org.opensearch.knn.common.KNNConstants.HNSW_ALGO_M;
 import static org.opensearch.knn.common.KNNConstants.KNN_ENGINE;
 import static org.opensearch.knn.common.KNNConstants.METHOD_HNSW;
 import static org.opensearch.knn.common.KNNConstants.METHOD_PARAMETER_EF_CONSTRUCTION;
 import static org.opensearch.knn.common.KNNConstants.METHOD_PARAMETER_M;
-import static org.opensearch.knn.common.KNNConstants.METHOD_PARAMETER_SPACE_TYPE;
 
 /**
  * Utility class for KNNVectorFieldMapper
@@ -121,15 +118,6 @@ static void validateIfCircuitBreakerIsNotTriggered() {
         }
     }
 
-    /**
-     * Validate if plugin is enabled
-     */
-    static void validateIfKNNPluginEnabled() {
-        if (!KNNSettings.isKNNPluginEnabled()) {
-            throw new IllegalStateException("KNN plugin is disabled. To enable update knn.plugin.enabled setting to true");
-        }
-    }
-
     /**
      * Prerequisite: Index should a knn index which is validated via index settings index.knn setting. This function
      * assumes that caller has already validated that index is a KNN index.
@@ -156,51 +144,12 @@ static boolean useFullFieldNameValidation(final Version indexCreatedVersion) {
         return indexCreatedVersion != null && indexCreatedVersion.onOrAfter(Version.V_2_17_0);
     }
 
-    public static SpaceType getSpaceType(final Settings indexSettings) {
-        String spaceType = indexSettings.get(KNNSettings.INDEX_KNN_SPACE_TYPE.getKey());
-        if (spaceType == null) {
-            spaceType = KNNSettings.INDEX_KNN_DEFAULT_SPACE_TYPE;
-            log.info(
-                String.format(
-                    "[KNN] The setting \"%s\" was not set for the index. Likely caused by recent version upgrade. Setting the setting to the default value=%s",
-                    METHOD_PARAMETER_SPACE_TYPE,
-                    spaceType
-                )
-            );
-        }
-        return SpaceType.getSpace(spaceType);
-    }
-
     private static int getM(Settings indexSettings) {
-        String m = indexSettings.get(KNNSettings.INDEX_KNN_ALGO_PARAM_M_SETTING.getKey());
-        if (m == null) {
-            log.info(
-                String.format(
-                    "[KNN] The setting \"%s\" was not set for the index. Likely caused by recent version upgrade. Setting the setting to the default value=%s",
-                    HNSW_ALGO_M,
-                    KNNSettings.INDEX_KNN_DEFAULT_ALGO_PARAM_M
-                )
-            );
-            return KNNSettings.INDEX_KNN_DEFAULT_ALGO_PARAM_M;
-        }
-        return Integer.parseInt(m);
+        return KNNSettings.INDEX_KNN_DEFAULT_ALGO_PARAM_M;
     }
 
     private static int getEfConstruction(Settings indexSettings, Version indexVersion) {
-        final String efConstruction = indexSettings.get(KNNSettings.INDEX_KNN_ALGO_PARAM_EF_CONSTRUCTION_SETTING.getKey());
-        if (efConstruction == null) {
-            final int defaultEFConstructionValue = IndexHyperParametersUtil.getHNSWEFConstructionValue(indexVersion);
-            log.info(
-                String.format(
-                    "[KNN] The setting \"%s\" was not set for the index. Likely caused by recent version upgrade. "
-                        + "Picking up default value for the index =%s",
-                    HNSW_ALGO_EF_CONSTRUCTION,
-                    defaultEFConstructionValue
-                )
-            );
-            return defaultEFConstructionValue;
-        }
-        return Integer.parseInt(efConstruction);
+        return IndexHyperParametersUtil.getHNSWEFConstructionValue(indexVersion);
     }
 
     static KNNMethodContext createKNNMethodContextFromLegacy(

src/main/java/org/opensearch/knn/index/query/KNNQuery.java

@@ -21,7 +21,6 @@
 import org.apache.lucene.search.Weight;
 import org.apache.lucene.search.join.BitSetProducer;
 import org.opensearch.common.StopWatch;
-import org.opensearch.knn.index.KNNSettings;
 import org.opensearch.knn.index.VectorDataType;
 import org.opensearch.knn.index.query.rescore.RescoreContext;
 
@@ -171,9 +170,6 @@ public KNNQuery filterQuery(Query filterQuery) {
      */
     @Override
     public Weight createWeight(IndexSearcher searcher, ScoreMode scoreMode, float boost) throws IOException {
-        if (!KNNSettings.isKNNPluginEnabled()) {
-            throw new IllegalStateException("KNN plugin is disabled. To enable update knn.plugin.enabled to true");
-        }
         StopWatch stopWatch = null;
         if (log.isDebugEnabled()) {
             stopWatch = new StopWatch().start();

src/main/java/org/opensearch/knn/jni/FaissService.java

@@ -407,19 +407,6 @@ public static native KNNQueryResult[] queryBinaryIndexWithFilter(
      */
     public static native byte[] trainByteIndex(Map<String, Object> indexParameters, int dimension, long trainVectorsPointer);
 
-    /**
-     * <p>
-     * The function is deprecated. Use {@link JNICommons#storeVectorData(long, float[][], long)}
-     * </p>
-     * Transfer vectors from Java to native
-     *
-     * @param vectorsPointer pointer to vectors in native memory. Should be 0 to create vector as well
-     * @param trainingData data to be transferred
-     * @return pointer to native memory location of training data
-     */
-    @Deprecated(since = "2.14.0", forRemoval = true)
-    public static native long transferVectors(long vectorsPointer, float[][] trainingData);
-
     /**
      * Range search index with filter
      *

src/main/java/org/opensearch/knn/jni/JNIService.java

@@ -419,21 +419,6 @@ public static byte[] trainIndex(Map<String, Object> indexParameters, int dimensi
         );
     }
 
-    /**
-     * <p>
-     * The function is deprecated. Use {@link JNICommons#storeVectorData(long, float[][], long, boolean)}
-     * </p>
-     * Transfer vectors from Java to native
-     *
-     * @param vectorsPointer pointer to vectors in native memory. Should be 0 to create vector as well
-     * @param trainingData   data to be transferred
-     * @return pointer to native memory location of training data
-     */
-    @Deprecated(since = "2.14.0", forRemoval = true)
-    public static long transferVectors(long vectorsPointer, float[][] trainingData) {
-        return FaissService.transferVectors(vectorsPointer, trainingData);
-    }
-
     /**
      * Range search index for a given query vector
      *

src/main/java/org/opensearch/knn/training/FloatTrainingDataConsumer.java

@@ -15,7 +15,6 @@
 import org.opensearch.action.search.SearchResponse;
 import org.opensearch.knn.index.engine.qframe.QuantizationConfig;
 import org.opensearch.knn.jni.JNICommons;
-import org.opensearch.knn.jni.JNIService;
 import org.opensearch.knn.index.memory.NativeMemoryAllocation;
 import org.opensearch.knn.quantization.factory.QuantizerFactory;
 import org.opensearch.knn.quantization.models.quantizationOutput.BinaryQuantizationOutput;
@@ -59,9 +58,10 @@ public void accept(List<?> floats) {
             }
         } else {
             trainingDataAllocation.setMemoryAddress(
-                JNIService.transferVectors(
+                JNICommons.storeVectorData(
                     trainingDataAllocation.getMemoryAddress(),
-                    floats.stream().map(v -> ArrayUtils.toPrimitive((Float[]) v)).toArray(float[][]::new)
+                    floats.stream().map(v -> ArrayUtils.toPrimitive((Float[]) v)).toArray(float[][]::new),
+                    floats.size()
                 )
             );
         }

src/test/java/org/opensearch/knn/index/KNNCreateIndexFromModelTests.java

@@ -21,6 +21,7 @@
 import org.opensearch.knn.index.engine.MethodComponentContext;
 import org.opensearch.knn.index.mapper.CompressionLevel;
 import org.opensearch.knn.index.mapper.Mode;
+import org.opensearch.knn.jni.JNICommons;
 import org.opensearch.knn.jni.JNIService;
 import org.opensearch.knn.index.engine.KNNEngine;
 import org.opensearch.knn.indices.Model;
@@ -49,7 +50,7 @@ public void testCreateIndexFromModel() throws IOException, InterruptedException
         int dimension = 3;
 
         // "Train" a faiss flat index - this really just creates an empty index that does brute force k-NN
-        long vectorsPointer = JNIService.transferVectors(0, new float[0][0]);
+        long vectorsPointer = JNICommons.storeVectorData(0, new float[0][0], 0);
         byte[] modelBlob = JNIService.trainIndex(
             ImmutableMap.of(INDEX_DESCRIPTION_PARAMETER, "Flat", SPACE_TYPE, spaceType.getValue()),
             dimension,