crystaldiskmark-parser 1.0.1

Read in the .txt files generated by CrystalDiskMark in Python

CrystalDiskMark-Parser

Read in the .txt files generated by CrystalDiskMark in Python. This library was created to enable visualizing the benchmark results of CrystalDiskMark in an automated way. The following plot created with matplotlib is an example.

Install

pip install crystaldiskmark-parser

See https://pypi.org/project/crystaldiskmark-parser/ for more information.

Overview

The library provides two Functions:

• parse(filepath): Returns parsed data as a crystaldiskmark_parser.BenchmarkResult
• parse_df(filepath): Returns parsed data as a pandas.DataFrame

Using parse_df(filepath) is recommended.

Example 0: Print DataFrame

from crystaldiskmark_parser.parser import parse_df

df = parse_df("../test/data/CrystalDiskMark_20210622162528 WD Blue 3D 1TB WDS100T2B0A.txt")
print(df)

Output

                  date                              test  \
0  2021/06/22 17:19:21  1 GiB (x5) [E: 96% (894/932GiB)]   
1  2021/06/22 17:19:21  1 GiB (x5) [E: 96% (894/932GiB)]   
2  2021/06/22 17:19:21  1 GiB (x5) [E: 96% (894/932GiB)]   
3  2021/06/22 17:19:21  1 GiB (x5) [E: 96% (894/932GiB)]   
4  2021/06/22 17:19:21  1 GiB (x5) [E: 96% (894/932GiB)]   
5  2021/06/22 17:19:21  1 GiB (x5) [E: 96% (894/932GiB)]   
6  2021/06/22 17:19:21  1 GiB (x5) [E: 96% (894/932GiB)]   
7  2021/06/22 17:19:21  1 GiB (x5) [E: 96% (894/932GiB)]   

                             time                                    os  \
0  Measure 5 sec / Interval 5 sec  Windows 10  [10.0 Build 19042] (x64)   
1  Measure 5 sec / Interval 5 sec  Windows 10  [10.0 Build 19042] (x64)   
2  Measure 5 sec / Interval 5 sec  Windows 10  [10.0 Build 19042] (x64)   
3  Measure 5 sec / Interval 5 sec  Windows 10  [10.0 Build 19042] (x64)   
4  Measure 5 sec / Interval 5 sec  Windows 10  [10.0 Build 19042] (x64)   
5  Measure 5 sec / Interval 5 sec  Windows 10  [10.0 Build 19042] (x64)   
6  Measure 5 sec / Interval 5 sec  Windows 10  [10.0 Build 19042] (x64)   
7  Measure 5 sec / Interval 5 sec  Windows 10  [10.0 Build 19042] (x64)   

      mode  profile         comment read_write type  blocksize unit_blocksize  \
0  [Admin]  Default  WD Blue 3D 1TB       read  SEQ        1.0            MiB   
1  [Admin]  Default  WD Blue 3D 1TB       read  SEQ        1.0            MiB   
2  [Admin]  Default  WD Blue 3D 1TB       read  RND        4.0            KiB   
3  [Admin]  Default  WD Blue 3D 1TB       read  RND        4.0            KiB   
4  [Admin]  Default  WD Blue 3D 1TB      write  SEQ        1.0            MiB   
5  [Admin]  Default  WD Blue 3D 1TB      write  SEQ        1.0            MiB   
6  [Admin]  Default  WD Blue 3D 1TB      write  RND        4.0            KiB   
7  [Admin]  Default  WD Blue 3D 1TB      write  RND        4.0            KiB   

  queues threads     rate unit_rate     iops unit_iops   latency unit_latency  
0      8       1  531.458      MB/s    506.8      IOPS  15726.77           us  
1      1       1  489.838      MB/s    467.1      IOPS   2139.41           us  
2     32       1  269.406      MB/s  65772.9      IOPS    470.58           us  
3      1       1   41.375      MB/s  10101.3      IOPS     98.75           us  
4      8       1  498.656      MB/s    475.6      IOPS  16750.30           us  
5      1       1  485.163      MB/s    462.7      IOPS   2159.18           us  
6     32       1  222.691      MB/s  54367.9      IOPS    583.02           us  
7      1       1   99.900      MB/s  24389.6      IOPS     40.73           us  

Example 1: Visualize benchmark results

from crystaldiskmark_parser.parser import parse_df
import matplotlib.pyplot as plt

#  Read to one big DataFrame
df = parse_df("../test/data/CrystalDiskMark_20210622162528 WD Blue 3D 1TB WDS100T2B0A.txt")
df = df.append(parse_df("../test/data/CrystalDiskMark_20210622163451 SAMSUNG 840 EVO 120GB.txt"))
df = df.append(parse_df("../test/data/CrystalDiskMark_20210626154221 Crucial P2 2TB on H87M.txt"))
df = df.append(parse_df("../test/data/CrystalDiskMark_20210627172807 Crucial P2 2TB on X370.txt"))
print(df.info())

# Extract relevant rows
data = df.loc[(df['read_write'] == 'read')
              & (df['type'] == "SEQ")
              & (df['queues'] == 8)
              & (df['blocksize'] == 1)
              & (df['unit_blocksize'] == "MiB")]

# Extract relevant columns
rate = data["rate"]
name = data["comment"]

# Plot
fig, ax = plt.subplots(figsize=(10, 6))
con = ax.bar(name, rate)
ax.bar_label(con)
ax.set_title("Sequential Read Rate of SSDs")
ax.set_ylabel("Rate [MB/s]")
plt.show()

Example 2: Visualize benchmark results (extended)

from crystaldiskmark_parser.parser import parse_df

import matplotlib.pyplot as plt


#%%
# Read to one big DataFrame
df = parse_df("../test/data/CrystalDiskMark_20210622162528 WD Blue 3D 1TB WDS100T2B0A.txt")
df = df.append(parse_df("../test/data/CrystalDiskMark_20210622163451 SAMSUNG 840 EVO 120GB.txt"))
df = df.append(parse_df("../test/data/CrystalDiskMark_20210626154221 Crucial P2 2TB on H87M.txt"))
df = df.append(parse_df("../test/data/CrystalDiskMark_20210627172807 Crucial P2 2TB on X370.txt"))

# Extract relevant rows
data = df.loc[(df['read_write'] == 'read')
              & (df['type'] == "SEQ")
              & (df['queues'] == 8)
              & (df['blocksize'] == 1)
              & (df['unit_blocksize'] == "MiB")]

# Extract relevant columns
read_rate = data["rate"]
read_rate = [round(x) for x in read_rate]
name = data["comment"]

# Extract relevant rows
data = df.loc[(df['read_write'] == 'write')
              & (df['type'] == "SEQ")
              & (df['queues'] == 8)
              & (df['blocksize'] == 1)
              & (df['unit_blocksize'] == "MiB")]

# Extract relevant columns
write_rate = data["rate"]
write_rate = [round(x) for x in write_rate]

# Plot
fig, ax = plt.subplots(figsize=(12, 6))
con = ax.bar([float(x)  + .25 for x in  range(len(read_rate))], read_rate, tick_label=name, width = 0.25)
ax.bar_label(con)
con = ax.bar([float(x)  + .5 for x in  range(len(write_rate))], write_rate, width = 0.25)
ax.bar_label(con)
ax.set_title("Sequential Read & Write Rate of SSDs")
ax.set_ylabel("Rate [MB/s]")
plt.legend(["read", "write"])
plt.show()

Example 3: Visualize benchmark results (without DataFrame)

from crystaldiskmark_parser.parser import parse
import matplotlib.pyplot as plt

result_wd = parse("data/CrystalDiskMark_20210622162528 WD Blue 3D 1TB WDS100T2B0A.txt")
result_samsung = parse("data/CrystalDiskMark_20210622163451 SAMSUNG 840 EVO 120GB.txt")

data = [result_wd.read_results[0].rate, result_samsung.read_results[0].rate]

fig, ax = plt.subplots(figsize=(10,6))
con = ax.bar(["WD Blue 3D 1TB", "Samsung 840 EVO 120GB"], data)
ax.bar_label(con)
ax.set_title("Read Rate of SSDs")
ax.set_ylabel(f"Rate [{result_wd.read_results[0].unit_rate}]")
plt.show()

Example 4: Print BenchmarkResult

from crystaldiskmark_parser.parser import parse

benchmark_result = parse("data/CrystalDiskMark_20210622162528 WD Blue 3D 1TB WDS100T2B0A.txt")
print(benchmark_result)

Output

BenchmarkResult({'test': '1 GiB (x5) [E: 96% (894/932GiB)]', 'date': '2021/06/22 17:19:21', 'os': 'Windows 10  [10.0 Build 19042] (x64)', 'profile': 'Default', 'time': 'Measure 5 sec / Interval 5 sec', 'mode': '[Admin]', 'comment': 'WD Blue 3D 1TB WDS100T2B0A', 'write_results': [TestResult({'test_type': 'SEQ', 'block_size': 1.0, 'unit_block_size': 'MiB', 'queues': 8, 'threads': 1, 'rate': 498.656, 'unit_rate': 'MB/s', 'iops': 475.6, 'unit_iops': 'IOPS', 'latency': 16750.3, 'unit_latency': 'us'}), TestResult({'test_type': 'SEQ', 'block_size': 1.0, 'unit_block_size': 'MiB', 'queues': 1, 'threads': 1, 'rate': 485.163, 'unit_rate': 'MB/s', 'iops': 462.7, 'unit_iops': 'IOPS', 'latency': 2159.18, 'unit_latency': 'us'}), TestResult({'test_type': 'RND', 'block_size': 4.0, 'unit_block_size': 'KiB', 'queues': 32, 'threads': 1, 'rate': 222.691, 'unit_rate': 'MB/s', 'iops': 54367.9, 'unit_iops': 'IOPS', 'latency': 583.02, 'unit_latency': 'us'}), TestResult({'test_type': 'RND', 'block_size': 4.0, 'unit_block_size': 'KiB', 'queues': 1, 'threads': 1, 'rate': 99.9, 'unit_rate': 'MB/s', 'iops': 24389.6, 'unit_iops': 'IOPS', 'latency': 40.73, 'unit_latency': 'us'})], 'read_results': [TestResult({'test_type': 'SEQ', 'block_size': 1.0, 'unit_block_size': 'MiB', 'queues': 8, 'threads': 1, 'rate': 531.458, 'unit_rate': 'MB/s', 'iops': 506.8, 'unit_iops': 'IOPS', 'latency': 15726.77, 'unit_latency': 'us'}), TestResult({'test_type': 'SEQ', 'block_size': 1.0, 'unit_block_size': 'MiB', 'queues': 1, 'threads': 1, 'rate': 489.838, 'unit_rate': 'MB/s', 'iops': 467.1, 'unit_iops': 'IOPS', 'latency': 2139.41, 'unit_latency': 'us'}), TestResult({'test_type': 'RND', 'block_size': 4.0, 'unit_block_size': 'KiB', 'queues': 32, 'threads': 1, 'rate': 269.406, 'unit_rate': 'MB/s', 'iops': 65772.9, 'unit_iops': 'IOPS', 'latency': 470.58, 'unit_latency': 'us'}), TestResult({'test_type': 'RND', 'block_size': 4.0, 'unit_block_size': 'KiB', 'queues': 1, 'threads': 1, 'rate': 41.375, 'unit_rate': 'MB/s', 'iops': 10101.3, 'unit_iops': 'IOPS', 'latency': 98.75, 'unit_latency': 'us'})]})

Development

Build and Publish a New Release

This repository has a CI/CD pipeline that handle building and publishing the package to PyPI. It is triggered by tagging a commit. The pipeline publishes only tagged commits on PyPI.

Steps to publish a new release:

1. Check if the version number is correctly set in setup.py and __init__.py
2. Create a new Release in GitHub. Set the version number as the name
3. Wait until pipeline ends
4. Check if package is available on PyPI