vectordb-bench 0.0.21__py3-none-any.whl → 0.0.23__py3-none-any.whl

vectordb_bench/backend/clients/__init__.py

@@ -38,8 +38,11 @@ class DB(Enum):
     Chroma = "Chroma"
     AWSOpenSearch = "OpenSearch"
     AliyunElasticsearch = "AliyunElasticsearch"
+    MariaDB = "MariaDB"
     Test = "test"
     AliyunOpenSearch = "AliyunOpenSearch"
+    MongoDB = "MongoDB"
+    TiDB = "TiDB"
 
     @property
     def init_cls(self) -> type[VectorDB]:  # noqa: PLR0911, PLR0912, C901
@@ -129,6 +132,21 @@ class DB(Enum):
 
             return AliyunOpenSearch
 
+        if self == DB.MongoDB:
+            from .mongodb.mongodb import MongoDB
+
+            return MongoDB
+
+        if self == DB.MariaDB:
+            from .mariadb.mariadb import MariaDB
+
+            return MariaDB
+
+        if self == DB.TiDB:
+            from .tidb.tidb import TiDB
+
+            return TiDB
+
         if self == DB.Test:
             from .test.test import Test
 
@@ -225,6 +243,21 @@ class DB(Enum):
 
             return AliyunOpenSearchConfig
 
+        if self == DB.MongoDB:
+            from .mongodb.config import MongoDBConfig
+
+            return MongoDBConfig
+
+        if self == DB.MariaDB:
+            from .mariadb.config import MariaDBConfig
+
+            return MariaDBConfig
+
+        if self == DB.TiDB:
+            from .tidb.config import TiDBConfig
+
+            return TiDBConfig
+
         if self == DB.Test:
             from .test.config import TestConfig
 
@@ -302,6 +335,21 @@ class DB(Enum):
 
             return AliyunOpenSearchIndexConfig
 
+        if self == DB.MongoDB:
+            from .mongodb.config import MongoDBIndexConfig
+
+            return MongoDBIndexConfig
+
+        if self == DB.MariaDB:
+            from .mariadb.config import _mariadb_case_config
+
+            return _mariadb_case_config.get(index_type)
+
+        if self == DB.TiDB:
+            from .tidb.config import TiDBIndexConfig
+
+            return TiDBIndexConfig
+
         # DB.Pinecone, DB.Chroma, DB.Redis
         return EmptyDBCaseConfig
 

vectordb_bench/backend/clients/api.py

@@ -25,6 +25,7 @@ class IndexType(str, Enum):
     ES_HNSW = "hnsw"
     ES_IVFFlat = "ivfflat"
     GPU_IVF_FLAT = "GPU_IVF_FLAT"
+    GPU_BRUTE_FORCE = "GPU_BRUTE_FORCE"
     GPU_IVF_PQ = "GPU_IVF_PQ"
     GPU_CAGRA = "GPU_CAGRA"
     SCANN = "scann"

vectordb_bench/backend/clients/aws_opensearch/aws_opensearch.py

@@ -12,6 +12,7 @@ log = logging.getLogger(__name__)
 
 WAITING_FOR_REFRESH_SEC = 30
 WAITING_FOR_FORCE_MERGE_SEC = 30
+SECONDS_WAITING_FOR_REPLICAS_TO_BE_ENABLED_SEC = 30
 
 
 class AWSOpenSearch(VectorDB):
@@ -52,10 +53,27 @@ class AWSOpenSearch(VectorDB):
         return AWSOpenSearchIndexConfig
 
     def _create_index(self, client: OpenSearch):
+        cluster_settings_body = {
+            "persistent": {
+                "knn.algo_param.index_thread_qty": self.case_config.index_thread_qty,
+                "knn.memory.circuit_breaker.limit": self.case_config.cb_threshold,
+            }
+        }
+        client.cluster.put_settings(cluster_settings_body)
         settings = {
             "index": {
                 "knn": True,
+                "number_of_shards": self.case_config.number_of_shards,
+                "number_of_replicas": 0,
+                "translog.flush_threshold_size": self.case_config.flush_threshold_size,
+                # Setting trans log threshold to 5GB
+                **(
+                    {"knn.algo_param.ef_search": self.case_config.ef_search}
+                    if self.case_config.engine == AWSOS_Engine.nmslib
+                    else {}
+                ),
             },
+            "refresh_interval": self.case_config.refresh_interval,
         }
         mappings = {
             "properties": {
@@ -145,9 +163,9 @@ class AWSOpenSearch(VectorDB):
                 docvalue_fields=[self.id_col_name],
                 stored_fields="_none_",
             )
-            log.info(f"Search took: {resp['took']}")
-            log.info(f"Search shards: {resp['_shards']}")
-            log.info(f"Search hits total: {resp['hits']['total']}")
+            log.debug(f"Search took: {resp['took']}")
+            log.debug(f"Search shards: {resp['_shards']}")
+            log.debug(f"Search hits total: {resp['hits']['total']}")
             return [int(h["fields"][self.id_col_name][0]) for h in resp["hits"]["hits"]]
         except Exception as e:
             log.warning(f"Failed to search: {self.index_name} error: {e!s}")
@@ -157,12 +175,37 @@ class AWSOpenSearch(VectorDB):
         """optimize will be called between insertion and search in performance cases."""
         # Call refresh first to ensure that all segments are created
         self._refresh_index()
-        self._do_force_merge()
+        if self.case_config.force_merge_enabled:
+            self._do_force_merge()
+            self._refresh_index()
+        self._update_replicas()
         # Call refresh again to ensure that the index is ready after force merge.
         self._refresh_index()
         # ensure that all graphs are loaded in memory and ready for search
         self._load_graphs_to_memory()
 
+    def _update_replicas(self):
+        index_settings = self.client.indices.get_settings(index=self.index_name)
+        current_number_of_replicas = int(index_settings[self.index_name]["settings"]["index"]["number_of_replicas"])
+        log.info(
+            f"Current Number of replicas are {current_number_of_replicas}"
+            f" and changing the replicas to {self.case_config.number_of_replicas}"
+        )
+        settings_body = {"index": {"number_of_replicas": self.case_config.number_of_replicas}}
+        self.client.indices.put_settings(index=self.index_name, body=settings_body)
+        self._wait_till_green()
+
+    def _wait_till_green(self):
+        log.info("Wait for index to become green..")
+        while True:
+            res = self.client.cat.indices(index=self.index_name, h="health", format="json")
+            health = res[0]["health"]
+            if health != "green":
+                break
+            log.info(f"The index {self.index_name} has health : {health} and is not green. Retrying")
+            time.sleep(SECONDS_WAITING_FOR_REPLICAS_TO_BE_ENABLED_SEC)
+        log.info(f"Index {self.index_name} is green..")
+
     def _refresh_index(self):
         log.debug(f"Starting refresh for index {self.index_name}")
         while True:
@@ -179,6 +222,12 @@ class AWSOpenSearch(VectorDB):
         log.debug(f"Completed refresh for index {self.index_name}")
 
     def _do_force_merge(self):
+        log.info(f"Updating the Index thread qty to {self.case_config.index_thread_qty_during_force_merge}.")
+
+        cluster_settings_body = {
+            "persistent": {"knn.algo_param.index_thread_qty": self.case_config.index_thread_qty_during_force_merge}
+        }
+        self.client.cluster.put_settings(cluster_settings_body)
         log.debug(f"Starting force merge for index {self.index_name}")
         force_merge_endpoint = f"/{self.index_name}/_forcemerge?max_num_segments=1&wait_for_completion=false"
         force_merge_task_id = self.client.transport.perform_request("POST", force_merge_endpoint)["task"]

vectordb_bench/backend/clients/aws_opensearch/cli.py

@@ -18,6 +18,79 @@ class AWSOpenSearchTypedDict(TypedDict):
     port: Annotated[int, click.option("--port", type=int, default=443, help="Db Port")]
     user: Annotated[str, click.option("--user", type=str, default="admin", help="Db User")]
     password: Annotated[str, click.option("--password", type=str, help="Db password")]
+    number_of_shards: Annotated[
+        int,
+        click.option("--number-of-shards", type=int, help="Number of primary shards for the index", default=1),
+    ]
+    number_of_replicas: Annotated[
+        int,
+        click.option(
+            "--number-of-replicas", type=int, help="Number of replica copies for each primary shard", default=1
+        ),
+    ]
+    index_thread_qty: Annotated[
+        int,
+        click.option(
+            "--index-thread-qty",
+            type=int,
+            help="Thread count for native engine indexing",
+            default=4,
+        ),
+    ]
+
+    index_thread_qty_during_force_merge: Annotated[
+        int,
+        click.option(
+            "--index-thread-qty-during-force-merge",
+            type=int,
+            help="Thread count during force merge operations",
+            default=4,
+        ),
+    ]
+
+    number_of_indexing_clients: Annotated[
+        int,
+        click.option(
+            "--number-of-indexing-clients",
+            type=int,
+            help="Number of concurrent indexing clients",
+            default=1,
+        ),
+    ]
+
+    number_of_segments: Annotated[
+        int,
+        click.option("--number-of-segments", type=int, help="Target number of segments after merging", default=1),
+    ]
+
+    refresh_interval: Annotated[
+        int,
+        click.option(
+            "--refresh-interval", type=str, help="How often to make new data available for search", default="60s"
+        ),
+    ]
+
+    force_merge_enabled: Annotated[
+        int,
+        click.option("--force-merge-enabled", type=bool, help="Whether to perform force merge operation", default=True),
+    ]
+
+    flush_threshold_size: Annotated[
+        int,
+        click.option(
+            "--flush-threshold-size", type=str, help="Size threshold for flushing the transaction log", default="5120mb"
+        ),
+    ]
+
+    cb_threshold: Annotated[
+        int,
+        click.option(
+            "--cb-threshold",
+            type=str,
+            help="k-NN Memory circuit breaker threshold",
+            default="50%",
+        ),
+    ]
 
 
 class AWSOpenSearchHNSWTypedDict(CommonTypedDict, AWSOpenSearchTypedDict, HNSWFlavor2): ...
@@ -36,6 +109,17 @@ def AWSOpenSearch(**parameters: Unpack[AWSOpenSearchHNSWTypedDict]):
             user=parameters["user"],
             password=SecretStr(parameters["password"]),
         ),
-        db_case_config=AWSOpenSearchIndexConfig(),
+        db_case_config=AWSOpenSearchIndexConfig(
+            number_of_shards=parameters["number_of_shards"],
+            number_of_replicas=parameters["number_of_replicas"],
+            index_thread_qty=parameters["index_thread_qty"],
+            number_of_segments=parameters["number_of_segments"],
+            refresh_interval=parameters["refresh_interval"],
+            force_merge_enabled=parameters["force_merge_enabled"],
+            flush_threshold_size=parameters["flush_threshold_size"],
+            number_of_indexing_clients=parameters["number_of_indexing_clients"],
+            index_thread_qty_during_force_merge=parameters["index_thread_qty_during_force_merge"],
+            cb_threshold=parameters["cb_threshold"],
+        ),
         **parameters,
     )

vectordb_bench/backend/clients/aws_opensearch/config.py

@@ -39,6 +39,16 @@ class AWSOpenSearchIndexConfig(BaseModel, DBCaseConfig):
     efConstruction: int = 256
     efSearch: int = 256
     M: int = 16
+    index_thread_qty: int | None = 4
+    number_of_shards: int | None = 1
+    number_of_replicas: int | None = 0
+    number_of_segments: int | None = 1
+    refresh_interval: str | None = "60s"
+    force_merge_enabled: bool | None = True
+    flush_threshold_size: str | None = "5120mb"
+    number_of_indexing_clients: int | None = 1
+    index_thread_qty_during_force_merge: int
+    cb_threshold: str | None = "50%"
 
     def parse_metric(self) -> str:
         if self.metric_type == MetricType.IP:

vectordb_bench/backend/clients/mariadb/cli.py

@@ -0,0 +1,107 @@
+from typing import Annotated, Optional, Unpack
+
+import click
+import os
+from pydantic import SecretStr
+
+from ....cli.cli import (
+    CommonTypedDict,
+    HNSWFlavor1,
+    cli,
+    click_parameter_decorators_from_typed_dict,
+    run,
+)
+from vectordb_bench.backend.clients import DB
+
+
+class MariaDBTypedDict(CommonTypedDict):
+    user_name: Annotated[
+        str, click.option("--username",
+                          type=str,
+                          help="Username",
+                          required=True,
+                          ),
+    ]
+    password: Annotated[
+        str, click.option("--password",
+                     type=str,
+                     help="Password",
+                     required=True,
+                     ),
+    ]
+
+    host: Annotated[
+        str, click.option("--host",
+                          type=str,
+                          help="Db host",
+                          default="127.0.0.1",
+                          ),
+    ]
+
+    port: Annotated[
+        int, click.option("--port",
+                          type=int,
+                          default=3306,
+                          help="Db Port",
+                          ),
+    ]
+
+    storage_engine: Annotated[
+        int, click.option("--storage-engine",
+                          type=click.Choice(["InnoDB", "MyISAM"]),
+                          help="DB storage engine",
+                          required=True,
+                          ),
+    ]
+
+class MariaDBHNSWTypedDict(MariaDBTypedDict):
+    ...
+    m: Annotated[
+        Optional[int], click.option("--m",
+                                    type=int,
+                                    help="M parameter in MHNSW vector indexing",
+                                    required=False,
+                                    ),
+    ]
+
+    ef_search: Annotated[
+        Optional[int], click.option("--ef-search",
+                                    type=int,
+                                    help="MariaDB system variable mhnsw_min_limit",
+                                    required=False,
+                                    ),
+    ]
+
+    max_cache_size: Annotated[
+        Optional[int], click.option("--max-cache-size",
+                                    type=int,
+                                    help="MariaDB system variable mhnsw_max_cache_size",
+                                    required=False,
+                                    ),
+    ]
+
+
+@cli.command()
+@click_parameter_decorators_from_typed_dict(MariaDBHNSWTypedDict)
+def MariaDBHNSW(
+    **parameters: Unpack[MariaDBHNSWTypedDict],
+):
+    from .config import MariaDBConfig, MariaDBHNSWConfig
+
+    run(
+        db=DB.MariaDB,
+        db_config=MariaDBConfig(
+            db_label=parameters["db_label"],
+            user_name=parameters["username"],
+            password=SecretStr(parameters["password"]),
+            host=parameters["host"],
+            port=parameters["port"],
+        ),
+        db_case_config=MariaDBHNSWConfig(
+            M=parameters["m"],
+            ef_search=parameters["ef_search"],
+            storage_engine=parameters["storage_engine"],
+            max_cache_size=parameters["max_cache_size"],
+        ),
+        **parameters,
+    )

vectordb_bench/backend/clients/mariadb/config.py

@@ -0,0 +1,71 @@
+from pydantic import SecretStr, BaseModel
+from typing import TypedDict
+from ..api import DBConfig, DBCaseConfig, MetricType, IndexType
+
+class MariaDBConfigDict(TypedDict):
+    """These keys will be directly used as kwargs in mariadb connection string,
+        so the names must match exactly mariadb API"""
+
+    user: str
+    password: str
+    host: str
+    port: int
+
+
+class MariaDBConfig(DBConfig):
+    user_name: str = "root"
+    password: SecretStr
+    host: str = "127.0.0.1"
+    port: int = 3306
+
+    def to_dict(self) -> MariaDBConfigDict:
+        pwd_str = self.password.get_secret_value()
+        return {
+            "host": self.host,
+            "port": self.port,
+            "user": self.user_name,
+            "password": pwd_str,
+        }
+
+
+class MariaDBIndexConfig(BaseModel):
+    """Base config for MariaDB"""
+
+    metric_type: MetricType | None = None
+
+    def parse_metric(self) -> str:
+        if self.metric_type == MetricType.L2:
+            return "euclidean"
+        elif self.metric_type == MetricType.COSINE:
+            return "cosine"
+        else:
+            raise ValueError(f"Metric type {self.metric_type} is not supported!")
+
+class MariaDBHNSWConfig(MariaDBIndexConfig, DBCaseConfig):
+    M: int | None
+    ef_search: int | None
+    index: IndexType = IndexType.HNSW
+    storage_engine: str = "InnoDB"
+    max_cache_size: int | None
+
+    def index_param(self) -> dict:
+        return {
+            "storage_engine": self.storage_engine,
+            "metric_type": self.parse_metric(),
+            "index_type": self.index.value,
+            "M": self.M,
+            "max_cache_size": self.max_cache_size,
+        }
+
+    def search_param(self) -> dict:
+        return {
+            "metric_type": self.parse_metric(),
+            "ef_search": self.ef_search,
+        }
+
+
+_mariadb_case_config = {
+        IndexType.HNSW: MariaDBHNSWConfig,
+}
+
+

vectordb_bench/backend/clients/mariadb/mariadb.py

@@ -0,0 +1,214 @@
+from ..api import VectorDB
+
+import logging
+from contextlib import contextmanager
+from typing import Any, Optional, Tuple
+from ..api import VectorDB
+from .config import MariaDBConfigDict, MariaDBIndexConfig
+import numpy as np
+
+import mariadb
+
+log = logging.getLogger(__name__)
+
+class MariaDB(VectorDB):
+    def __init__(
+            self,
+            dim: int,
+            db_config: MariaDBConfigDict,
+            db_case_config: MariaDBIndexConfig,
+            collection_name: str = "vec_collection",
+            drop_old: bool = False,
+            **kwargs,
+        ):
+
+        self.name = "MariaDB"
+        self.db_config = db_config
+        self.case_config = db_case_config
+        self.db_name = "vectordbbench"
+        self.table_name = collection_name
+        self.dim = dim
+
+        # construct basic units
+        self.conn, self.cursor = self._create_connection(**self.db_config)
+        
+        if drop_old:
+            self._drop_db()
+            self._create_db_table(dim)
+
+        self.cursor.close()
+        self.conn.close()
+        self.cursor = None
+        self.conn = None
+
+
+    @staticmethod
+    def _create_connection(**kwargs) -> Tuple[mariadb.Connection, mariadb.Cursor]:
+        conn = mariadb.connect(**kwargs)
+        cursor = conn.cursor()
+
+        assert conn is not None, "Connection is not initialized"
+        assert cursor is not None, "Cursor is not initialized"
+
+        return conn, cursor
+
+
+    def _drop_db(self):
+        assert self.conn is not None, "Connection is not initialized"
+        assert self.cursor is not None, "Cursor is not initialized"
+        log.info(f"{self.name} client drop db : {self.db_name}")
+
+        # flush tables before dropping database to avoid some locking issue
+        self.cursor.execute("FLUSH TABLES")
+        self.cursor.execute(f"DROP DATABASE IF EXISTS {self.db_name}")
+        self.cursor.execute("COMMIT")
+        self.cursor.execute("FLUSH TABLES")
+
+    def _create_db_table(self, dim: int):
+        assert self.conn is not None, "Connection is not initialized"
+        assert self.cursor is not None, "Cursor is not initialized"
+
+        index_param = self.case_config.index_param()
+
+        try:
+            log.info(f"{self.name} client create database : {self.db_name}")
+            self.cursor.execute(f"CREATE DATABASE {self.db_name}")
+
+            log.info(f"{self.name} client create table : {self.table_name}")
+            self.cursor.execute(f"USE {self.db_name}")
+
+            self.cursor.execute(f"""
+              CREATE TABLE {self.table_name} (
+                id INT PRIMARY KEY,
+                v VECTOR({self.dim}) NOT NULL
+              ) ENGINE={index_param["storage_engine"]}
+            """)
+            self.cursor.execute("COMMIT")
+
+        except Exception as e:
+            log.warning(
+                f"Failed to create table: {self.table_name} error: {e}"
+            )
+            raise e from None
+
+
+    @contextmanager
+    def init(self) -> None:
+        """ create and destory connections to database.
+
+        Examples:
+            >>> with self.init():
+            >>>     self.insert_embeddings()
+        """
+        self.conn, self.cursor = self._create_connection(**self.db_config)
+
+        index_param = self.case_config.index_param()
+        search_param = self.case_config.search_param()
+
+        # maximize allowed package size
+        self.cursor.execute("SET GLOBAL max_allowed_packet = 1073741824")
+
+        if index_param["index_type"] == "HNSW":
+            if index_param["max_cache_size"] != None:
+                self.cursor.execute(f"SET GLOBAL mhnsw_max_cache_size = {index_param["max_cache_size"]}")
+            if search_param["ef_search"] != None:
+                self.cursor.execute(f"SET mhnsw_ef_search = {search_param["ef_search"]}")
+            self.cursor.execute("COMMIT")
+
+        self.insert_sql = f"INSERT INTO {self.db_name}.{self.table_name} (id, v) VALUES (%s, %s)"
+        self.select_sql = f"SELECT id FROM {self.db_name}.{self.table_name} ORDER by vec_distance_{search_param["metric_type"]}(v, %s) LIMIT %d"
+        self.select_sql_with_filter = f"SELECT id FROM {self.db_name}.{self.table_name} WHERE id >= %d ORDER by vec_distance_{search_param["metric_type"]}(v, %s) LIMIT %d"
+
+        try:
+            yield
+        finally:
+            self.cursor.close()
+            self.conn.close()
+            self.cursor = None
+            self.conn = None
+    
+
+    def ready_to_load(self) -> bool:
+        pass
+
+    def optimize(self) -> None:
+        assert self.conn is not None, "Connection is not initialized"
+        assert self.cursor is not None, "Cursor is not initialized"
+
+        index_param = self.case_config.index_param()
+
+        try:
+            index_options = f"DISTANCE={index_param['metric_type']}"
+            if index_param["index_type"] == "HNSW" and index_param["M"] != None:
+                index_options += f" M={index_param['M']}"
+
+            self.cursor.execute(f"""
+              ALTER TABLE {self.db_name}.{self.table_name}
+              ADD VECTOR KEY v(v) {index_options}
+            """)
+            self.cursor.execute("COMMIT")
+
+        except Exception as e:
+            log.warning(
+                f"Failed to create index: {self.table_name} error: {e}"
+            )
+            raise e from None
+
+        pass
+
+    @staticmethod
+    def vector_to_hex(v):
+        return np.array(v, 'float32').tobytes()
+
+    def insert_embeddings(
+        self,
+        embeddings: list[list[float]],
+        metadata: list[int],
+        **kwargs: Any,
+    ) -> Tuple[int, Optional[Exception]]:
+        """Insert embeddings into the database.
+        Should call self.init() first.
+        """
+        assert self.conn is not None, "Connection is not initialized"
+        assert self.cursor is not None, "Cursor is not initialized"
+
+        try:
+            metadata_arr = np.array(metadata)
+            embeddings_arr = np.array(embeddings)
+
+            batch_data = []
+            for i, row in enumerate(metadata_arr):
+                batch_data.append((int(row), self.vector_to_hex(embeddings_arr[i])));
+
+            self.cursor.executemany(self.insert_sql, batch_data)
+            self.cursor.execute("COMMIT")
+            self.cursor.execute("FLUSH TABLES")
+
+            return len(metadata), None
+        except Exception as e:
+            log.warning(
+                f"Failed to insert data into Vector table ({self.table_name}), error: {e}"
+            )
+            return 0, e
+    
+
+    def search_embedding(
+        self,
+        query: list[float],
+        k: int = 100,
+        filters: dict | None = None,
+        timeout: int | None = None,
+        **kwargs: Any,
+    ) -> (list[int]):
+        assert self.conn is not None, "Connection is not initialized"
+        assert self.cursor is not None, "Cursor is not initialized"
+
+        search_param = self.case_config.search_param()
+
+        if filters:
+            self.cursor.execute(self.select_sql_with_filter, (filters.get('id'), self.vector_to_hex(query), k))
+        else:
+            self.cursor.execute(self.select_sql, (self.vector_to_hex(query), k))
+
+        return [id for id, in self.cursor.fetchall()]
+