feat(runner): add configurable --serial-cooldown between concurrent a…

README.md

@@ -424,7 +424,8 @@ Execute tests for the index types: HGraph.
 NUM_PER_BATCH=10000 vectordbbench hologreshgraph --host Hologres_Endpoint --port 80 \
 --user ACCESS_ID --password ACCESS_KEY --database DATABASE_NAME \
 --m 64 --ef-construction 400 --case-type Performance768D10M \
---index-type HGraph --ef-search 400 --k 10 --num-concurrency 1,60,70,75,80,90,95,100,110,120
+--index-type HGraph --ef-search 400 --k 10 --num-concurrency 1,60,70,75,80,90,95,100,105,110,115,120,125,130 \
+--serial-cooldown 3
 ```
 
 To list the options for Hologres, execute `vectordbbench hologreshgraph --help`, The following are some Hologres-specific command-line options.

vectordb_bench/__init__.py

@@ -35,6 +35,7 @@ class config:
     CONCURRENCY_DURATION = 30
 
     CONCURRENCY_TIMEOUT = 3600
+    SERIAL_COOLDOWN = 0
     CLOUD_INSERT_READINESS_TIMEOUT = env.float("CLOUD_INSERT_READINESS_TIMEOUT", None)
     CLOUD_INSERT_READINESS_POLL_INTERVAL = env.float("CLOUD_INSERT_READINESS_POLL_INTERVAL", 5.0)
 

vectordb_bench/backend/task_runner.py

@@ -318,6 +318,12 @@ def _run_perf_case(self, drop_old: bool = True) -> Metric:
                         m.conc_latency_avg_list,
                     ) = search_results
                 if TaskStage.SEARCH_SERIAL in self.config.stages:
+                    cooldown = self.config.case_config.concurrency_search_config.serial_cooldown
+                    if TaskStage.SEARCH_CONCURRENT in self.config.stages and cooldown > 0:
+                        log.info(
+                            f"Cooldown {cooldown}s before serial search to ensure a stable measurement environment"
+                        )
+                        time.sleep(cooldown)
                     search_results = self._serial_search()
                     m.recall, m.ndcg, m.serial_latency_p99, m.serial_latency_p95 = search_results
             m.payload_profile = self.ca.payload_profile.value

vectordb_bench/cli/cli.py

@@ -390,6 +390,16 @@ class CommonTypedDict(TypedDict):
             "Set to a negative value to wait indefinitely.",
         ),
     ]
+    serial_cooldown: Annotated[
+        float,
+        click.option(
+            "--serial-cooldown",
+            type=float,
+            default=config.SERIAL_COOLDOWN,
+            show_default=True,
+            help="Cooldown in seconds between concurrent and serial search phases",
+        ),
+    ]
     custom_case_name: Annotated[
         str,
         click.option(
@@ -805,6 +815,7 @@ def run(
                 concurrency_duration=parameters["concurrency_duration"],
                 num_concurrency=[int(s) for s in parameters["num_concurrency"]],
                 concurrency_timeout=parameters["concurrency_timeout"],
+                serial_cooldown=parameters["serial_cooldown"],
             ),
             custom_case=get_custom_case_config(parameters),
         ),

vectordb_bench/models.py

@@ -184,6 +184,7 @@ class ConcurrencySearchConfig(BaseModel):
     num_concurrency: list[int] = config.NUM_CONCURRENCY
     concurrency_duration: int = config.CONCURRENCY_DURATION
     concurrency_timeout: int = config.CONCURRENCY_TIMEOUT
+    serial_cooldown: float = config.SERIAL_COOLDOWN
 
 
 class CaseConfig(BaseModel):