Updated (12 April) - Boosting your production scheduling performance (Part 2)

Update: 12 April 2022 -- We sincerely apologize. We not just published the below blog post but also published a new version of just plan it.

We recognized that we failed with our testing procedures. We also recognized that the new version does not meet our quality standards and expectations. As a consequence, we today rolled back our just plan it software to the status quo that we had before. 

We continue to work on the below behind the scenes and still aim at gaining the below-mentioned performance increases. We will let you know once they will be rolled out again - matching the quality standards that we missed this time.

Again: My sincere apologies (Elmar K.).

With this release of our production scheduling software for job shops, just plan it, we finalize our current development focus on increasing software performance.
We put a lot of thought and effort into enhancing jpi for all our users and strongly believe that the numbers we can present today were worth the effort.
This also means that we refocus now on adding and enhancing functionalities and delivering new and useful ways for your planning. But, of course, this shift in focus does not mean, that we stop keeping an eye on our stability and performance, which are still important to us.

The first quarter of 2022

We started this year coming off a big release in December 2021, which allowed us to offer a lot of exciting new functionalities like pdf exports, the world view, and more. Changing our fundamental visualization engine of course needed some additional love in January, but we still developed new functionality in the form of the AutoRelease Dates. But we also started to acknowledge that we need to work on our performance, as we saw that with the complexities and features introduced in the last years and the growth of our customers' plans, we saw a decline in response times, which we set out to reduce as much as possible.

This led to us rethinking architecture and communication systems, as well as implementing intelligence to our backend to decide if changes have a small enough impact to forgo rescheduling and just updating the data model. We released the first step of this in February, including not only the update to the backend but also preparing the front-end client to reduce its load, as, especially for larger plans, updating the visualization is the most time-consuming part of an update. This first step, took the same differentiation into account as the backend update, speeding up non-scheduling relevant changes.

But of course, we wanted to expand this to as many types of updates to your schedule as possible. And this is where we are. Using the new foundation we built in December we are now able to significantly increase client performance. Let’s take a closer look at the last 3 months' worth of efforts:

The results

Let me preface this: All results are measured against the January 31, 2022 release of just plan it.
To check if we are on the right track with what we are doing, we established a baseline of interactions and datasets of different sizes and complexities to measure against.

We focused our measurements on the OnClient duration, meaning the time just plan it takes to render and visualize the schedule to you, as except for the differentiation between non-scheduling relevant and scheduling-relevant changes, we did not have that many avenues to increase our backend calculation at this time.

This means that the values presented will not represent the total time an interaction takes, but only the amount of time of the visualization. Of course, all measurements were taken on the same PC to keep them comparable.

With these ground rules out of the way let's take a look at the improvements of non-scheduling-relevant changes:

In this table, you can see an example of our measurements. All interactions tested are represented by three columns: The performance end of January, the performance with the new release, and the percent improvement. The numbers given are milliseconds.
As you can see in most cases we more than halved the time spend on visualization and in some of the larger datasets we were even able to quarter it.

Overall we increased the visualization performance for non-scheduling-relevant changes by roughly 65 % of overall changes and datasets.

While this itself is already a good result, as I alluded to earlier we did not want to focus on non-scheduling-relevant changes and did our best to improve the scheduling-relevant changes as well:

And that we did. As the nature of scheduling-relevant changes usually needs to update a significant portion of the schedule, the performance increase is not quite as significant as with the non-scheduling-relevant changes, but I would say an average increase in performance of about 41% is nothing to scoff at. Also as you might have guessed the larger and more complicated your schedule is, the higher the potential performance increase will be.

We are proud of what we have been able to achieve with this release, increasing the visualization performance on average between all datasets and interactions by roughly 53%, and are looking forward to your feedback and now refocusing on adding new functionality. 

What is next

The next sprint will focus on adding the first step into the foray of an often requested feature: Team planning

We will also start the groundwork for another longer-term project we will reveal at a later time.

And of course, as always we will include bug fixes and smaller enhancements.

We plan to release the next sprint on May 9, 2022.

That’s it for today. This release is important for us, as it really helped us see new possibilities and showed us new paths forward to improve just plan it, to help you improve your scheduling.

As always we are happy and eager to receive any feedback from you guys and to answer any questions you may have.

Until then,

Happy Scheduling!

Further information

We continuously enhance our software. Want to learn more about our releases? Check this website, listing product news and updates.