Zipfian movie/theater IDs (#413)

* Optionally pick zipfian movie and theater ids

* Compare zipfian IDs to bounds

* Add command line flags

* Use store_true action

---------

Co-authored-by: Geoffrey Yu <[email protected]>

workloads/IMDB_extended/run_transactions.py

@@ -42,7 +42,12 @@ def noop_handler(_signal, _frame):
     signal.signal(signal.SIGINT, noop_handler)
 
     worker = TransactionWorker(
-        worker_idx, args.seed ^ worker_idx, args.scale_factor, args.dataset_type
+        worker_idx,
+        args.seed ^ worker_idx,
+        args.scale_factor,
+        args.dataset_type,
+        args.use_zipfian_ids,
+        args.zipfian_alpha,
     )
 
     txn_prng = random.Random(~(args.seed ^ worker_idx))
@@ -281,6 +286,19 @@ def main():
         help="This controls the range of reads the transaction worker performs, "
         "depending on the dataset size.",
     )
+    parser.add_argument(
+        "--use-zipfian-ids",
+        action="store_true",
+        help="Whether the transaction worker should draw movie and theatre IDs "
+        "from a Zipfian distribution.",
+    )
+    parser.add_argument(
+        "--zipfian-alpha",
+        type=float,
+        default=1.1,
+        help="The alpha parameter for the Zipfian distribution. Only used if "
+        "--use-zipfian-ids is `True`. Must be strictly greater than 1. ",
+    )
     # These three arguments are used for the day long experiment.
     parser.add_argument(
         "--num-client-path",

workloads/IMDB_extended/workload_utils/transaction_worker.py

@@ -1,4 +1,5 @@
 import random
+import numpy as np
 import logging
 from datetime import datetime, timedelta
 from typing import List, Tuple, Any
@@ -26,9 +27,15 @@ def __init__(
         seed: int,
         scale_factor: int,
         dataset_type: str = "original",
+        use_zipfian_ids: bool = False,
+        zipfian_alpha: float = 1.1,
     ) -> None:
         self.worker_id = worker_id
         self.prng = random.Random(seed)
+        self.use_zipfian_ids = use_zipfian_ids
+        if use_zipfian_ids:
+            self.zprng = np.random.default_rng(seed)
+            self.zipfian_alpha = zipfian_alpha
 
         self.min_movie_id = MIN_MOVIE_ID
         if dataset_type == "original":
@@ -50,6 +57,15 @@ def __init__(
         self.loc_max = 1e6
         self.showing_years = 2
 
+    def _sample_id(self, min_id: int, max_id: int) -> int:
+        sampled = (
+            self.zprng.zipf(self.zipfian_alpha) - 1 + min_id
+            if self.use_zipfian_ids
+            else self.prng.randint(min_id, max_id)
+        )
+
+        return min(max_id, sampled)
+
     def edit_movie_note(self, db: Database) -> bool:
         """
         Represents editing the "misc info" for a specific movie.
@@ -64,7 +80,7 @@ def edit_movie_note(self, db: Database) -> bool:
         """
 
         # 1. Select a random movie id.
-        movie_id = self.prng.randint(self.min_movie_id, self.max_movie_id)
+        movie_id = self._sample_id(self.min_movie_id, self.max_movie_id)
 
         try:
             # Start the transaction.
@@ -114,10 +130,10 @@ def add_new_showing(self, db: Database) -> bool:
         - Insert into showing
         """
         # 1. Select a random theatre id.
-        theatre_id = self.prng.randint(self.min_theatre_id, self.max_theatre_id)
+        theatre_id = self._sample_id(self.min_theatre_id, self.max_theatre_id)
 
         # 2. Select a random movie id.
-        movie_id = self.prng.randint(self.min_movie_id, self.max_movie_id)
+        movie_id = self._sample_id(self.min_movie_id, self.max_movie_id)
 
         showings_to_add = self.prng.randint(*self.showings_to_add)
 
@@ -167,7 +183,7 @@ def purchase_tickets(self, db: Database, select_using_name: bool) -> bool:
         """
 
         # 1. Select a random theatre number.
-        theatre_num = self.prng.randint(self.min_theatre_id, self.max_theatre_id)
+        theatre_num = self._sample_id(self.min_theatre_id, self.max_theatre_id)
 
         try:
             # Start the transaction.