This tool is a realtime data visualisation tool for FIO written using PyQtGraph.
This application is written in Python, using PyQtGraph to graph FIO data in realtime. Running as root is required because this application accesses block devices.
- README.md
- This readme file
- LICENSE.txt
- License file
- fio_visualizer.py
- The frontend for this application
- realtime_back.py
- The backend for this application
- mainwindow.ui
- GUI definition part of the script
- Workloads:
- Device
- Block device workloads covering typical specifications, separated in SATA and NVMe sections
- Precondition
- instruction and script files to prepare SSD for the testing.
- Simulation
- Workloads to simulate real applications.
- MultiDevice
- Workloads to test two or more drives at a time. Currently has limitted support due to improper handling of total performance in the script.
- Device
Requires:
- FIO 2.15 or newer (2.1.x is no longer supported) found here
- PyQtGraph found here
- Python 2.7 or 3+ here
- PyQt 4.8+ or PySide
- numpy
- scipy
-
You should have python and PyQt installed with this distro
-
Install pyqtgraph-develop (0.9.9 required) from here
$ python setup.py install
-
Install Cyphon from here Version 0.21 or higher is required.
$ python setup.py install
-
Install Numpy from here
$ python setup.py build$ python setup.py install
-
Install FIO from here
# ./configure# make# make install
-
Run Visualizer as root.
# ./fio_visualizer.py
-
Add the following repository
deb http://luke.campagnola.me/debian dev/into your/etc/apt/sources.listand then run$ sudo apt-get install fio python-pyqtgraph -
Install FIO from here
# ./configure# make# make install
-
Run Visualizer as root.
# ./fio_visualizer.py
- Increased CPU load with numjobs > 4 and all threads plotting. Avoid plotting all threads with numjobs > 4 or disable unneccessary ones to save CPU resources.
- Multi-jobs configurations files (i.e. [job1]... [job2]) are not supported at the moment. The total amount of jobs if every workload has "numjobs" specified too are not correctly parsed. This can still be implemented manually specifing it in the code.
Continue working on workload profiles for new Intel SSDs.
- Andrey Kudryavtsev andrey.o.kudryavtsev@intel.com
- Alexey Ponomarev alexey.ponomarev@intel.com