update README, docker-compose.yaml and change performance graph colours

README.md

@@ -7,7 +7,7 @@ This repo contains a tool for benchmarking Weaviate performance.
 * 📊 results and context can be found in the [Weaviate documentation](https://weaviate.io/developers/weaviate/current/benchmarks/)
 * 💬 discuss the results on our [Slack channel](https://join.slack.com/t/weaviate/shared_invite/zt-goaoifjr-o8FuVz9b1HLzhlUfyfddhw) or [Twitter](https://twitter.com/weaviate_io)
 
-### ANN benchmark
+## ANN benchmark
 
 There are two components you will need to run for the benchmarks:
 
@@ -16,42 +16,30 @@ There are two components you will need to run for the benchmarks:
 
 You can run both as containers on the same machine via Docker compose.
 
-For replicating our benchmarks we recommend setting up two separate machines.
+For replicating our benchmarks we recommend setting the following machine:
 
-| Machine description | CPU type | CPUs | Memory | Disk size | Disk type | Misc. |
+| Machine name | CPU type | CPUs | Memory | Disk size | Disk type | Misc. |
 | --- | --- | --- | --- | --- | --- | --- |
-| Machine to run Weaviate | c2 | 30 | 120GB | 500GB | SSD | [Ubuntu 22.04 with Docker-compose](https://gist.github.com/bobvanluijt/04f6d97916244a7de59fead84ef63cd4) |
-| Machine to run benchmark script | N2 | 8 | 64GB | 500GB | SSD | [Ubuntu 22.04 with Docker-compose](https://gist.github.com/bobvanluijt/04f6d97916244a7de59fead84ef63cd4) |
+| `n4-highmem-16` | N4 | 16 | 128GB | 512GB | Hyperdisk Balanced | Debian 12 (bookworm) with [Docker and Compose V2](https://gist.github.com/StefanBogdan/821d18bbc5f18978643adff508749cf0) |
 
-#### Prepare the Weaviate machine
+### Run tests
 
-Clone this repo and cd into it `$ git clone https://github.com/semi-technologies/weaviate-benchmarking && cd weaviate-benchmarking`
-
-Run the following command to spin up Weaviate: `$ docker-compose up weaviate -d`
-
-Copy the IP address and amount of CPU cores this machine has.
-
-#### Prepare the benchmark machine
-
-Check if the Weaviate machine is available: `curl http://{IP OF WEAVIATE INSTANCE}/v1/meta`. Note that the instance runs on port `8080`, e.g., `http://10.128.15.12:8080/v1/meta`.
-You will also need to allow port `50051` for gRPC and `21121` for metrics. You can verify this via `nc -zv localhost 50051`.
-
-Clone this repo and cd into it `$ git clone https://github.com/semi-technologies/weaviate-benchmarking && cd weaviate-benchmarking`
+Clone this repo and cd into it `$ git clone https://github.com/weaviate/weaviate-benchmarking && cd weaviate-benchmarking`
 
 Download the files into a datasets folder as outlined below.
 
 ```sh
 mkdir datasets && \
-    curl -o ./datasets/dbpedia-100k-openai-ada002-angular.hdf5 https://storage.googleapis.com/ann-datasets/custom/dbpedia-100k-openai-ada002-angular.hdf5 \
-    curl -o ./datasets/deep-image-96-angular.hdf5 https://ann-benchmarks.com/deep-image-96-angular.hdf5 && \
-    curl -o ./datasets/mnist-784-euclidean.hdf5 https://ann-benchmarks.com/mnist-784-euclidean.hdf5 && \
-    curl -o ./datasets/gist-960-euclidean.hdf5 https://ann-benchmarks.com/gist-960-euclidean.hdf5
+    curl -o ./datasets/dbpedia-openai-1000k-angular.hdf5 https://storage.googleapis.com/ann-datasets/ann-benchmarks/dbpedia-openai-1000k-angular.hdf5 && \
+    curl -o ./datasets/snowflake-msmarco-arctic-embed-m-v1.5-angular.hdf5 https://storage.googleapis.com/ann-datasets/custom/snowflake-msmarco-arctic-embed-m-v1.5-angular.hdf5 && \
+    curl -o ./datasets/sift-128-euclidean.hdf5 http://ann-benchmarks.com/sift-128-euclidean.hdf5 && \
+    curl -o ./datasets/sphere-10M-meta-dpr.hdf5 https://storage.googleapis.com/ann-datasets/custom/sphere-10M-meta-dpr.hdf5
 ```
 
 Run a single performance test on an [ann-benchmarks](https://ann-benchmarks.com/) hdf5 dataset.
 
 ```sh
-DATASET=./datasets/dbpedia-100k-openai-ada002-angular.hdf5 DISTANCE=cosine docker compose run benchmarker
+DATASET=./datasets/dbpedia-openai-1000k-angular.hdf5 DISTANCE=cosine docker compose up
 ```
 
 For more details on additional configuration options see the help options.

benchmarker/scripts/python/performance-graphs.py

@@ -4,8 +4,6 @@
 import glob
 import json
 import argparse
-import seaborn as sns
-import matplotlib.ticker as tkr
 import matplotlib.pyplot as plt
 import pandas as pd
 
@@ -40,8 +38,8 @@ def create_plot(results_df: pd.DataFrame, mode='light'):
 
     # Set custom colors for limits
     color_map = {
-        100: '#098f73',
-        10: '#2b17e7'
+        100: '#61bd73',
+        10: '#fc3988'
     }
 
     # Configure plot style based on mode

docker-compose.yml

@@ -7,8 +7,8 @@ services:
       dockerfile: Dockerfile
     command: >
       /app/benchmarker ann-benchmark
-      -v ${DATASET:-./datasets/dbpedia-100k-openai-ada002-angular.hdf5 }
-      -d ${DISTANCE:-cosine}
+      --vectors ${DATASET:-./datasets/dbpedia-100k-openai-ada002-angular.hdf5 }
+      --distance ${DISTANCE:-cosine}
       --grpcOrigin ${GRPC_ORIGIN:-weaviate:50051}
       --httpOrigin ${HTTP_ORIGIN:-weaviate:8080}
     volumes:
@@ -22,7 +22,7 @@ services:
     - '8080'
     - --scheme
     - http
-    image: docker.io/semitechnologies/weaviate:1.25.8
+    image: docker.io/semitechnologies/weaviate:1.27.1
     ports:
       - 8080:8080
       - 50051:50051