After converting an app from Mx6.6.0 to 6.10.3 (running on premise, windows with sql server db), users experience a slow performance. Simple overview pages and the upload/download/delete can take minutes. A testrobot is executing basic functionality in the app (login,upload,search,delete) to gain performance statistics. The statistics point out that the performance of uploading a file (using dropzone) is reduced from 1250ms to 6000ms exactly after converting to Mx6.10.3. Also deleting a file is slower (3500ms to 7000ms). The performance of the other basic functionalities stay stable, so my conclusion is that the something has changed in the filehandling in Mendix that causes to slow performance of the app. The release notes between Mx6.6.0 and Mx6.10.3 only mention performance improvements.... (buying and implementing an performance tool will not be a customer's choice) Does anyone has experience and advice with this problem? Edit: The console does not show strange behaviour during uploading multiple documents via Dropzone. The app is being moved from on-premise to Mendix cloud and there is a good chance this solved to performance issue's.
I think you should have a closer look at the DropZone Widget. Is the the basic upload of a single file using the standard widget (not the dropzone) also slow? Did you have a look at the browser console if there is a data traffic issue ?
I believe the way that the Mendix UI communicates with the server runtime changed in this span between 6.6 and 6.10, specifically in 6.8 (see these release notes). Specifically, I believe it sends changes made to objects in the browser alongside microflow calls or other server requests instead of sending them in a separate stream.
I wonder if this is an edge case where that actually decreases performance. Could you share a HAR file from the browser for your app in 6.6 and 6.10 so we can analyze the difference? This is done in the "network" tab of your browser's developer tools. Here's an article from Atlassian that overviews the general process of exporting a HAR file.
This should help us determine whether the problem is due to a change in network communication or a change within the runtime.
EDIT: By the way, if you have the opportunity to create a small test project, it would be great to submit that along with a support ticket.