Benchmarks

Maybe a bad idea.

Author: AdvancedImagingUTSW

docs/source/user_guide/file_formats.rst

@@ -36,10 +36,77 @@ to store metadata.
 ----------------
 
 BigDataViewer H5/N5/OME-Zarr
--------------------
+----------------------------
 
 **navigate** uses `h5py <https://docs.h5py.org/en/stable/index.html>`_ (H5) and
 `zarr <https://zarr.readthedocs.io/en/stable/>`_ (N5) to store data in a BigDataViewer
 file format. This is a pyramidal format, necessitating the saving of both the original
 data and down sampled versions of this data. The additional data slows down the write
 speed. The N5 format is faster than H5 because it allows multithreaded writes.
+
+----------------
+
+Image Writing Benchmarks
+------------------------
+
+To evaluate the performance of saving imaging data in different formats, we conducted
+benchmarks on a Linux-based system. We assessed the average disk write time for TIFF,
+OME-TIFF, H5, N5, and OME-Zarr formats across image resolutions of 512x512,
+1024x1024, and 2048x2048 under two conditions: (A) capturing 1000 single-plane
+images and (B) acquiring a single z-stack composed of 1000 planes. All times
+are measured in milliseconds. Bottom panel for both A and B has a reduced y-axis to
+highlight the differences in write times between the formats. Results are also
+presented in a table format below.
+
+The initial setup for writing H5/N5 files introduces
+significant overhead, and to a lesser extent for TIFF and OME-TIFF files, which
+elevates the average write time. However, the median write time remains consistently
+low and stable across most of the acquisition. For z-stack imaging, TIFF and OME-TIFF
+formats achieved write speeds of up to approximately 400 Hz for a 2048x2048 camera
+resolution, surpassing the operational speeds of most contemporary sCMOS cameras.
+The Big-TIFF variant was used for both TIFF and OME-TIFF formats to accommodate the
+large file sizes.
+
+
+.. image:: images/benchmarks.png
+    :width: 50%
+    :align: center
+
+.. table::
+   :widths: auto
+   :align: center
+
+   +-------------+---------+----------+-------+-------+---------+
+   | Z=1, T=1000 | TIFF    | OME-TIFF | H5    | N5    | OME-Zarr|
+   +=============+=========+==========+=======+=======+=========+
+   | 512x512     | 4.06    | 15.26    | 2.79  | 12.09 | 1.35    |
+   +-------------+---------+----------+-------+-------+---------+
+   | 1024x1024   | 6.69    | 15.25    | 8.33  | 14.74 | 2.46    |
+   +-------------+---------+----------+-------+-------+---------+
+   | 2048x2048   | 22.25   | 38.57    | 34.71 | 29.56 | 11.85   |
+   +-------------+---------+----------+-------+-------+---------+
+
+.. table::
+   :widths: auto
+   :align: center
+
+   +--------------+---------+----------+-------+-------+---------+
+   | Z=1000, T=1  | TIFF    | OME-TIFF | H5    | N5    | OME-Zarr|
+   +==============+=========+==========+=======+=======+=========+
+   | 512x512      | 0.21    | 0.17     | 4.76  | 1.80  | 1.30    |
+   +--------------+---------+----------+-------+-------+---------+
+   | 1024x1024    | 0.52    | 0.57     | 15.70 | 5.38  | 4.22    |
+   +--------------+---------+----------+-------+-------+---------+
+   | 2048x2048    | 2.35    | 2.14     | 75.56 | 14.08 | 8.60    |
+   +--------------+---------+----------+-------+-------+---------+
+
+The computer specifications that the benchmarks were performed on are as follows:
+
+-   Architecture: x86_64
+-   CPU operational modes: 32-bit, 64-bit
+-   Number of CPUs: 72
+-   Threads per core: 2
+-   Cores per socket: 18
+-   CPU Model: Intel(R) Xeon(R) Gold 6354 CPU @ 3.00GHz
+-   CPU Speed: 883.850 MHz
+-   Memory: 0.5 TB

docs/source/user_guide/hardware/supported_hardware.rst

@@ -3,7 +3,10 @@ Configuring Hardware Devices
 ============================
 
 The following section aims to provide users with the requisite information for configuring
-hardware devices. A high-level overview of the supported devices can be found :doc:`here <../hardware_overview>`.
+hardware devices. A high-level overview of the supported devices can be found
+:doc:`here <../hardware_overview>`. Additional devices are available by installing the
+**navigate-mmcore-plugin**. To learn more, please visit the **navigate-mmcore-plugin**
+`documentation <https://thedeanlab.github.io/navigate-mmcore-plugin/>`_.
 
 
 Data Acquisition Card