What is your MATLAB uptime

Someone was talking about running a process in MATLAB for a month: http://www.mathworks.com/matlabcentral/answers/21437-remote-monitoring-of-matlab-during-a-month-long-process

I have had simulations run for a few days and maybe have gone a week without restarting MATLAB.

What is your longest MATLAB uptime?

Answer 1

[Jan]

36 hours. It was an extreme programming session. The runtime of my programs is usually below a minute.

I shut down my machines every night. I think they need the sleep to get all the bits in the right order.

Answer 2

[Walter Roberson]

A month or so for me. I usually leave MATLAB open until I need to reboot my desktop to install a security update.

[Image Analyst]

But you weren't running just one m-file that whole time, right? I might have MATLAB running a couple of days or so at most before I decide to shutdown the computer to save energy. My individual m-files generally only take a minute or so to run.

Answer 3

[Sven]

I've got 2 machines with MATLAB always open :)

PC1. The desktop - hack around developing and testing.

PC2. The runner - Any task that would take more than 10 minutes I run on here.

Usually I'll be developing a new algorithm (searching/segmenting CTs) on PC1, and once I've got it stable enough to run without crashing for different input volumes I'll save my script/function/object, switch over to PC2 and let it iterate over a larger set of scans. My processes are usually short (< 10 min), but I loop them over scans sometimes for a day or two.

I hate it when I arrive in the morning to see my desktop machine rebooted with a windows update.

[Walter Roberson]

Back when I still had a working PC, I always turned off automatic Windows Update. Mind you, I did first get in to that habit in the days when the updates underwent less testing before release and so might crash (or corrupt) machines. (I have accumulated more good reasons since...)

Answer 4

[Kelly Kearney]

If my work computer is on, then Matlab is usually running; I'll usually keep a single session open for a week or so. And usually the session restart is due to either accumulated Java memory loss leading to graphics errors and a necessary restart, or a segmentation fault of some sort crashing things.

My longest single processes run for a few days; these involve a parent script running 50-100 iterations of a model (often in parallel), while a single model iteration runs for approximately 2-3 hours (integrating a system of ODEs). And as was noted in the thread you reference, I do write data to file as it's being calculated rather than trying to hold it in memory until the simulations finish... more memory efficient and it prevents data from being lost if the computer crashes.