Merge pull request #4 from chris-delg/docker

Docker

.dockerignore

@@ -0,0 +1,2 @@
+__pycache__/
+.vscode/

Dockerfile

@@ -0,0 +1,29 @@
+FROM continuumio/miniconda3
+
+# set working dir for docker app
+WORKDIR /app
+
+# install system dependencies including OpenGL
+RUN apt-get update && apt-get install -y \
+    python3-opengl \
+    mesa-utils \
+    libgl1-mesa-glx \
+    libgl1-mesa-dri
+
+# move environment.yml from source to docker app
+COPY environment.yml ./
+
+# installing dependancies
+RUN conda env create -n benchmark --file environment.yml
+
+# had source instead of conda here
+RUN echo "source activate benchmark" > ~/.bashrc 
+
+# set conda to path
+ENV PATH=/opt/conda/envs/benchmark/bin:$PATH
+
+# copy source code into docker app
+COPY . .
+
+# running the benchmark
+CMD [ "python", "main.py" ]

README.md

@@ -5,13 +5,25 @@ This benchmarking suite compares the following Zarr writers (click for documenta
 - [TensorStore](https://google.github.io/tensorstore/)
 - [Zarr-Python](https://zarr.readthedocs.io/en/stable/index.html)
 - [OME-Zarr](https://ome-zarr.readthedocs.io/en/stable/)
-- [z5](https://github.com/constantinpape/z5)
 
-To run the benchmark, just execute the following command.
-```bash 
+## Building and Running
+
+To build and run the benchmark in this program, you need to install the various Zarr writers on your machine. Refer below to see what options you have for your operating system.
+
+### Linux
+
+An "environment.yml" has been provided for simplifying the dependency process on Linux. Simply download [conda](https://conda.io/projects/conda/en/latest/user-guide/install/index.html) and execute the following commands to run the program.
+```bash
+conda env create -n benchmark --file environment.yml
+conda activate benchmark
 python main.py
 ```
 
-## Dependencies
+### Mac and Windows
 
-To build and run the benchmark in this program, you need to install the various Zarr writers on your machine. Refer to the documentation links above for installation instructions.
+Since some of the dependencies are only available on Linux, a Dockerfile has been left to allow building on other systems. Simply download [Docker](https://www.docker.com/) and execute the following commadns to run the program.
+
+```bash
+docker build -t benchmark .
+docker run -it benchmark
+```

environment.yml

@@ -0,0 +1,16 @@
+name: benchmark
+channels:
+  - defaults
+  - conda-forge
+  - rapidsai
+dependencies:
+  - pip
+  - matplotlib
+  - napari
+  - pyqt
+  - ome-zarr
+  - kvikio
+  - zarr=2.15*
+  - pip:
+      - ml-dtypes
+      - tensorstore

main.py

@@ -3,14 +3,14 @@
 
 
 def main() -> None:
+    bandwidth_map = {}
     shape = [64, 1080, 1920]
     chunks = [64, 540, 960]
     fig, graph = plt.subplots(2, 2)
 
     zarr_python = Zarr_Python(shape=shape, chunks=chunks)
     tensorstore = Tensorstore(shape=shape, chunks=chunks)
     ome_zarr = Ome_Zarr(shape=shape, chunks=chunks)
-    z5 = Z5(shape=shape, chunks=chunks)
 
     '''
     Append Tests:
@@ -19,8 +19,12 @@ def main() -> None:
         * TensorStore
         * Zarr Python
     '''
-    zarr_python.continuous_append_test(graph=graph[1][0], avg_graph=graph[1][1], append_dim_size=100)
-    tensorstore.continuous_append_test(graph=graph[1][0], avg_graph=graph[1][1], append_dim_size=100) 
+    bandwidth_map["TensorStore Append"] = (
+        tensorstore.continuous_append_test(graph=graph[1][0], avg_graph=graph[1][1], append_dim_size=400)
+    ) 
+    bandwidth_map["Zarr Python Append"] = (
+        zarr_python.continuous_append_test(graph=graph[1][0], avg_graph=graph[1][1], append_dim_size=400)
+    ) 
 
     # setting up graph for append tests
     graph[1][0].set_xlabel("Write Number")
@@ -34,12 +38,24 @@ def main() -> None:
         * TensorStore
         * Zarr Python
         * OME Zarr
-        * z5py
     '''
-    tensorstore.continuous_write_test(graph=graph[0][0], avg_graph=graph[0][1], append_dim_size=10, step=1)
-    zarr_python.continuous_write_test(graph=graph[0][0], avg_graph=graph[0][1], append_dim_size=10, step=1)
-    ome_zarr.continuous_write_test(graph=graph[0][0], avg_graph=graph[0][1], append_dim_size=10, step=1)   # crashes at anything above 16gigs (append_dim_size 90 on my machine)
-    z5.continuous_write_test(graph=graph[0][0], avg_graph=graph[0][1], append_dim_size=10, step=1)
+    bandwidth_map["TensorStore Write"] = (
+        tensorstore.continuous_write_test(graph=graph[0][0], avg_graph=graph[0][1], append_dim_size=80, step=5)
+    )   
+    bandwidth_map["Zarr Python Write"] = (
+        zarr_python.continuous_write_test(graph=graph[0][0], avg_graph=graph[0][1], append_dim_size=80, step=5)
+    )
+    # ome-zarr crashes at anything above 16gigs on my machine
+    bandwidth_map["OME Zarr Write"] = (
+        ome_zarr.continuous_write_test(graph=graph[0][0], avg_graph=graph[0][1], append_dim_size=45, step=5)
+    )      
+
+    # print the average bandwidth for each of the tests
+    print(f"Shape {shape}, Chunks {chunks}")
+    print("----------Bandwidth----------")
+    for test, bandwidth in bandwidth_map.items():
+        print(f"{test} : {bandwidth} GBps")
+    print("\n\n")
 
     # setting up graphs for write tests
     graph[0][0].set_xlabel("Data Size (GB)")

zarr_libraries/__init__.py

@@ -1,5 +1,4 @@
 from zarr_libraries.common import * 
 from zarr_libraries.tensorstore.tensorstore_zarr import *
 from zarr_libraries.zarr_python.zarr_python import *
-from zarr_libraries.ome_ngff.ome_zarr import *
-from zarr_libraries.z5py.z5 import *
+from zarr_libraries.ome_ngff.ome_zarr import *

zarr_libraries/ome_ngff/ome_zarr.py

@@ -45,7 +45,9 @@ def __continuous_write(self, result_path: str, append_dim_size: int, step: int)
         return file_sizes, bandwidths
 
 
-    def continuous_write_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: matplotlib.axes._axes.Axes, append_dim_size: int, step: int) -> None:
+    def continuous_write_test(self, graph: matplotlib.axes._axes.Axes, 
+                              avg_graph: matplotlib.axes._axes.Axes, 
+                              append_dim_size: int, step: int) -> int:
         # calls continuous write function and graphs results
         print("\n\n--------OME-Zarr Stress Test--------\n\n")
         file_sizes, bandwidths = self.__continuous_write(
@@ -56,5 +58,6 @@ def continuous_write_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: ma
         print("--------------------------------------------------------------\n\n")
         graph.plot(file_sizes, bandwidths, label="OME-Zarr", marker='o')
         avg_graph.bar("OME-Zarr", np.average(bandwidths))
+        return np.average(bandwidths)
 
 

zarr_libraries/tensorstore/tensorstore_zarr.py

@@ -114,7 +114,9 @@ def __continuous_append(self, result_path: str, append_dim_size: int) -> tuple[l
         return write_number, bandwidths
 
 
-    def continuous_write_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: matplotlib.axes._axes.Axes, append_dim_size: int, step: int) -> None:
+    def continuous_write_test(self, graph: matplotlib.axes._axes.Axes, 
+                              avg_graph: matplotlib.axes._axes.Axes, 
+                              append_dim_size: int, step: int) -> int:
         # calls continuous write function and graphs results
         print("\n\n--------Tensorstore Stress Test--------\n\n")
         file_sizes, bandwidths = self.__continuous_write(
@@ -125,9 +127,12 @@ def continuous_write_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: ma
         print("--------------------------------------------------------------\n\n")
         graph.plot(file_sizes, bandwidths, label="TensorStore", marker='o')
         avg_graph.bar("TensorStore", np.average(bandwidths))
+        return np.average(bandwidths)
 
 
-    def continuous_append_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: matplotlib.axes._axes.Axes, append_dim_size: int) -> None:
+    def continuous_append_test(self, graph: matplotlib.axes._axes.Axes, 
+                               avg_graph: matplotlib.axes._axes.Axes, 
+                               append_dim_size: int) -> int:
         # calls continuous append function and graphs results
         print("\n\n--------Tensorstore Stress Test--------\n\n")
         write_number, bandwidths = self.__continuous_append(
@@ -137,5 +142,6 @@ def continuous_append_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: m
         print("--------------------------------------------------------------\n\n")
         graph.plot(write_number, bandwidths, label="TensorStore")
         avg_graph.bar("TensorStore", np.average(bandwidths))
+        return np.average(bandwidths)
 
 

zarr_libraries/zarr_python/zarr_python.py

@@ -81,7 +81,9 @@ def __continuous_append(self, result_path: str, append_dim_size: int) -> tuple[l
         return write_number, bandwidths
 
 
-    def continuous_write_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: matplotlib.axes._axes.Axes, append_dim_size: int, step: int) -> None:
+    def continuous_write_test(self, graph: matplotlib.axes._axes.Axes, 
+                              avg_graph: matplotlib.axes._axes.Axes, 
+                              append_dim_size: int, step: int) -> int:
         # calls continuous write function and graphs results
         print("\n\n--------Zarr-Python Stress Test--------\n\n")
         file_sizes, bandwidths = self.__continuous_write(
@@ -92,9 +94,12 @@ def continuous_write_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: ma
         print("--------------------------------------------------------------\n\n")
         graph.plot(file_sizes, bandwidths, label="Zarr-Python", marker='o')
         avg_graph.bar("Zarr-Python", np.average(bandwidths))
-
+        return np.average(bandwidths)
+
 
-    def continuous_append_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: matplotlib.axes._axes.Axes, append_dim_size: int) -> None:
+    def continuous_append_test(self, graph: matplotlib.axes._axes.Axes, 
+                               avg_graph: matplotlib.axes._axes.Axes, 
+                               append_dim_size: int) -> int:
         # calls continuous append function and graphs results
         print("\n\n--------Zarr-Python Append Stress Test--------\n\n")
         write_number, bandwidths = self.__continuous_append(
@@ -104,5 +109,6 @@ def continuous_append_test(self, graph: matplotlib.axes._axes.Axes, avg_graph: m
         print("--------------------------------------------------------------\n\n")
         graph.plot(write_number, bandwidths, label="Zarr-Python")
         avg_graph.bar("Zarr-Python", np.average(bandwidths))
+        return np.average(bandwidths)