enable command line parameter

Author: slice4e

engine/base_client/client.py

@@ -4,6 +4,7 @@
 from pathlib import Path
 from typing import List, Dict, Any, Optional
 import warnings
+import random
 
 from benchmark import ROOT_DIR
 from benchmark.dataset import Dataset
@@ -174,7 +175,18 @@ def run_experiment(
         upload_end_idx: int = -1,
         num_queries: int = -1,
         ef_runtime: List[int] = [],
+        mixed_workload_params: dict = None,
     ):
+
+        # Extract mixed workload parameters
+        insert_fraction = 0.0
+        if mixed_workload_params:
+            insert_fraction = mixed_workload_params.get("insert_fraction", 0.1)
+            seed = mixed_workload_params.get("seed", None)
+            if seed is not None:
+                random.seed(seed)  # Set seed for reproducible patterns
+
+            print(f"In run_experiment. insert_fraction: {insert_fraction} seed: {seed}" )
         results = {"upload": {}, "search": {}}
         execution_params = self.configurator.execution_params(
             distance=dataset.config.distance, vector_size=dataset.config.vector_size
@@ -273,7 +285,7 @@ def run_experiment(
                     )
 
                     search_stats = searcher.search_all(
-                        dataset.config.distance, reader.read_queries(), num_queries
+                        dataset.config.distance, reader.read_queries(), num_queries, insert_fraction
                     )
                     # ensure we specify the client count in the results
                     search_params["parallel"] = client_count

engine/base_client/search.py

@@ -86,7 +86,8 @@ def search_all(
         num_queries: int = -1,
         insert_fraction: float = 0.0,
     ):
-        insert_fraction = self.search_params.get("insert_fraction", 0.0)
+        print(f"In base_client/search_all. insert_fraction {insert_fraction}")
+
         parallel = self.search_params.get("parallel", 1)
         top = self.search_params.get("top", None)
         single_search_params = self.search_params.get("search_params", None)
@@ -271,7 +272,6 @@ def process_chunk(chunk, search_one, insert_one, insert_fraction):
         if random.random() < insert_fraction:
             result = insert_one(query)
         else:
-            # Search
             result = search_one(query)
         results.append(result)
     return results

experiments/configurations/dbpedia-calibration-mixed.json

@@ -28,6 +28,10 @@
       "parallel": 100,
       "algorithm": "hnsw",
       "data_type": "FLOAT16"
+    },
+    "mixed_workload_params": {
+        "insert_fraction": 0.1,
+        "seed": 42
     }
   },
   {
@@ -59,6 +63,10 @@
       "parallel": 100,
       "algorithm": "hnsw",
       "data_type": "FLOAT32"
+    },
+    "mixed_workload_params": {
+      "insert_fraction": 0.1,
+      "seed": 42
     }
   },
 {
@@ -91,6 +99,10 @@
       "parallel": 100,
       "data_type": "FLOAT16",
       "algorithm": "svs-vamana"
+    },
+    "mixed_workload_params": {
+      "insert_fraction": 0.1,
+      "seed": 42
     }
   },
   {
@@ -123,6 +135,10 @@
       "parallel": 100,
       "data_type": "FLOAT32",
       "algorithm": "svs-vamana"
+    },
+    "mixed_workload_params": {
+      "insert_fraction": 0.1,
+      "seed": 42
     }
   },
   {
@@ -156,6 +172,10 @@
       "parallel": 100,
       "data_type": "FLOAT16",
       "algorithm": "svs-vamana"
+    },
+    "mixed_workload_params": {
+      "insert_fraction": 0.1,
+      "seed": 42
     }
   },
   {
@@ -189,6 +209,10 @@
       "parallel": 100,
       "data_type": "FLOAT32",
       "algorithm": "svs-vamana"
+    },
+    "mixed_workload_params": {
+      "insert_fraction": 0.1,
+      "seed": 42
     }
   },
   {
@@ -222,6 +246,10 @@
       "parallel": 100,
       "data_type": "FLOAT32",
       "algorithm": "svs-vamana"
+    },
+    "mixed_workload_params": {
+      "insert_fraction": 0.1,
+      "seed": 42
     }
   },
   {
@@ -255,6 +283,10 @@
       "parallel": 100,
       "data_type": "FLOAT16",
       "algorithm": "svs-vamana"
+    },
+    "mixed_workload_params": {
+      "insert_fraction": 0.1,
+      "seed": 42
     }
   },
   {
@@ -288,6 +320,10 @@
       "parallel": 100,
       "data_type": "FLOAT16",
       "algorithm": "svs-vamana"
+    },
+    "mixed_workload_params": {
+      "insert_fraction": 0.1,
+      "seed": 42
     }
   },
   {
@@ -321,6 +357,10 @@
       "parallel": 100,
       "data_type": "FLOAT32",
       "algorithm": "svs-vamana"
+    },
+    "mixed_workload_params": {
+      "insert_fraction": 0.1,
+      "seed": 42
     }
   }
 ]

run.py

@@ -30,10 +30,14 @@ def run(
     ef_runtime: List[int] = typer.Option([], help="Filter search experiments by ef runtime values. Only experiments with these ef values will be run."),
     describe: str = typer.Option(None, help="Describe available options: 'datasets' or 'engines'. When used, shows information and exits."),
     verbose: bool = typer.Option(False, "--verbose", "-v", help="Show detailed information when using --describe"),
+    mixed_workload: bool = typer.Option(False, help="Enable mixed workload mode"),
+    insert_fraction: float = typer.Option(0.1, help="Fraction of operations that are inserts (0.0-1.0)"),
+    mixed_workload_seed: int = typer.Option(None, help="Random seed for reproducible mixed workload patterns"),
 ):
     """
     Example:
         python3 run.py --engines *-m-16-* --engines qdrant-* --datasets glove-*
+        python3 run.py --engines redis --datasets glove-* --mixed-workload --insert-fraction 0.2
         python3 run.py --describe datasets
         python3 run.py --describe engines --verbose
     """
@@ -64,6 +68,14 @@ def run(
         if any(fnmatch.fnmatch(name, dataset) for dataset in datasets)
     }
 
+    mixed_params = {}
+    if mixed_workload:
+        mixed_params = {
+            "insert_fraction": insert_fraction,
+            "seed": mixed_workload_seed
+        }
+        print(f"Running mixed workload. insert_fraction: {insert_fraction} random_seed: {mixed_workload_seed}")
+
     for engine_name, engine_config in selected_engines.items():
         for dataset_name, dataset_config in selected_datasets.items():
             print(f"Running experiment: {engine_name} - {dataset_name}")
@@ -88,6 +100,7 @@ def run(
                         upload_end_idx,
                         queries,
                         ef_runtime,
+                        mixed_workload_params=mixed_params
                     )
                 client.delete_client()