Suggestion for Windows installations to reduce deployment times

8 views (last 30 days)
This is a product suggestion for the developers of all the MathWorks installers for Windows....
I work at a university where we deploy MATLAB to several departments' computer labs, instructor stations, and faculty and staff desktops. We use SCCM and create application packages to deploy to our endpoints. Because the Windows MATLAB installer contains thousands of mostly-small files, it takes a significant amount of time to just get the package "copied down" to endpoints before it actually kicks off the installation (setup.exe) process. Our most recent R2023a package includes around 40 or more products/toolboxes totaling 7GB in size with over 14,000 files. Because Windows (SMB and/or NTFS) is terribly inefficient at handling small files, this makes for a very slow deployment & installation process.
I'm proposing that all the MathWorks Windows installers be changed to support some kind of compressed archive format. This can be in the form of Installshield-type CAB files, standalone MSI installer files, or MSI installer files with separate sidecar <product/toolbox>.CAB files, or per-product/toolbox "ZIP" files. Ideally the installer process should be smart enough to extract the files from these archive files and save them directly to their final destination (install folder)--without going thru an intermediate step of extracting to a "temp" folder and then copying/deleting or moving them to their destination
While we can create workarounds such as building a self-extracting (SFX) ZIP/EXE file or create our own "WIM" image (similar to ISO) and "mount" it, NOT needing to do this would be a big time saver!
EDIT:
Also, please add some kind of "wait" command line option (or do this by default) so the installer will wait until all child installer processes are done before exiting the main setup.exe process. As it stands now, the main setup.exe process kicks off a child installer process and exits the main installer process. This makes deployment tools like SCCM think it finishes the installation within just a few seconds, which can be bad when you're staging multiple installs back-to-back or a reboot happens in the next step.
cheers!
Scott
  1 Comment
Scott Copus
Scott Copus on 3 Aug 2023
Edited: Scott Copus on 3 Aug 2023
In my testing, the Windows uninstall leaves behind (even after a restart) the C:\Program Files\MATLAB\R2023a\ folder with over 400 files of over 200MB. Does anyone know how to do a clean uninstall?
Here's the command I used:
"C:\Program Files\MATLAB\R2023a\uninstall\bin\win64\uninstall.exe" -inputFile uninstaller_input.txt
Does it work better when a separate temporary copy of the entire "uninstall" folder is used outside of the installation folder?

Sign in to comment.

Answers (1)

Walter Roberson
Walter Roberson on 3 Aug 2023

The matlab installer already uses archives.

Typically a Product consists of:

  • one archive for each architecture
  • one archive for shared files such as .m files and configuration files
  • one archive for documentation
  • archives having to do with localisation to different languages
  • a table of contents file
  1 Comment
Scott Copus
Scott Copus on 3 Aug 2023
Sorry, I should have clarified to say combined archive format, so as to reduce the total number of installation files... to make it more efficient on the underlying file system and also when transferring over a network.
For example, IBM's latest SPSS Statistics package is available in a single ~1GB MSI installer file that installs several packages including some open source packages totaling 1.7GB containing over 15K files once installed. Their installer also includes logic to automatically remove previous versions too. I think these design aspects are a golden standard and makes life much easier for enterprise app deployment admins... which would be great if MathWorks installers could start moving in that direction.

Sign in to comment.

Categories

Find more on Introduction to Installation and Licensing in Help Center and File Exchange

Products


Release

R2023a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!