From 80c67fdd73abc654519b4ce4f4236faa08ed44b6 Mon Sep 17 00:00:00 2001
From: Arthur Lu <learthurgo@gmail.com>
Date: Sun, 23 Feb 2025 19:17:55 +0000
Subject: [PATCH] add throughput and hit ratio metrics
---
app/config.yaml | 2 +-
tests/basic_latency_hitrate.py | 36 ----------------------
tests/random_readonly.py | 56 ++++++++++++++++++++++++++++++++++
3 files changed, 57 insertions(+), 37 deletions(-)
delete mode 100644 tests/basic_latency_hitrate.py
create mode 100644 tests/random_readonly.py
diff --git a/app/config.yaml b/app/config.yaml
index c8eb188..99494e9 100644
--- a/app/config.yaml
+++ b/app/config.yaml
@@ -1,4 +1,4 @@
-cache_strategy: "Baseline" # Change this to "Prefetch" or "Tiered" or "Seive"
+cache_strategy: "Tiered" # Change this to "Prefetch" or "Tiered" or "Seive"
cache_limit: 10
l2_cache_limit: 100
db_file: "llmData_sns.json" # Change this to the name of any json file within the "database/datastore" folder
\ No newline at end of file
diff --git a/tests/basic_latency_hitrate.py b/tests/basic_latency_hitrate.py
deleted file mode 100644
index 54aa8f5..0000000
--- a/tests/basic_latency_hitrate.py
+++ /dev/null
@@ -1,36 +0,0 @@
-# Tests latency and hit rate of endpoints. Can be configured with weighted averages for various endpoints.
-
-import requests
-import random
-import json
-
-baseurl = "http://localhost:8000"
-
-endpoints = {
- "/user/{user_id}": 1
-}
-
-user_ids = json.loads(requests.get(baseurl + "/users").content)["ids"]
-
-random.seed(0)
-
-def generate_random():
- x = random.choices(list(endpoints.keys()), list(endpoints.values()))[0] # select randomly from endpoint (keys) with weight (values)
-
- random_user = str(random.choice(user_ids))
- x = x.replace("{user_id}", random_user)
-
- return baseurl + x
-
-times = []
-hits = []
-
-for i in range(10000):
- url = generate_random()
- response = requests.get(url)
- content = json.loads(response.content)
- times.append(content["time_ms"])
- hits.append(content["source"] == "cache")
-
-print(f"average response time (ms): {sum(times) / len(times)}")
-print(f"hits: {sum(hits)} misses: {len(hits) - sum(hits)}")
\ No newline at end of file
diff --git a/tests/random_readonly.py b/tests/random_readonly.py
new file mode 100644
index 0000000..ab1b671
--- /dev/null
+++ b/tests/random_readonly.py
@@ -0,0 +1,56 @@
+# Tests latency and hit rate of endpoints. Can be configured with weighted averages for various endpoints.
+
+import requests
+import random
+import json
+from tqdm import tqdm
+import time
+
+baseurl = "http://localhost:8000"
+
+endpoints = {
+ "/user/{user_id}": 1
+}
+
+user_ids = json.loads(requests.get(baseurl + "/users").content)["ids"]
+
+random.seed(0)
+
+def generate_random():
+ x = random.choices(list(endpoints.keys()), list(endpoints.values()))[0] # select randomly from endpoint (keys) with weight (values)
+
+ random_user = str(random.choice(user_ids))
+ x = x.replace("{user_id}", random_user)
+
+ return baseurl + x
+
+times = []
+hits = []
+
+start = time.time()
+for i in tqdm(range(10000)):
+ url = generate_random()
+ response = requests.get(url)
+ content = json.loads(response.content)
+ times.append(content["time_ms"])
+ hits.append(content["source"] == "cache")
+end = time.time()
+
+hits_count = sum(hits)
+miss_count = len(hits) - hits_count
+
+hits_time = 0
+miss_time = 0
+for i in range(len(times)):
+ if hits[i]:
+ hits_time += times[i]
+ else:
+ miss_time += times[i]
+total_time = hits_time + miss_time
+
+print(f"hits: {hits_count} misses: {miss_count} ratio: { hits_count / (hits_count + miss_count)}")
+print(f"average response time (ms) : {total_time / len(times)}")
+print(f"average cache hit response time (ms) : {hits_time / hits_count}")
+print(f"average cache miss response time (ms): {miss_time / miss_count}")
+print(f"cache throughput (requests / ms) : { len(times) / total_time}")
+print(f"real throughput (requests / ms) : { len(times) / (end - start) / 1000}")
\ No newline at end of file