I'm asking only for a info. I have a problem, but it may be normal, just want to know. I deployed a flexdashboard app with shiny in runtime. I have this selectInput in which i can choose two datasets at the moment.
The fact is that is particularly slow in changing between the two datasets. Running on my pc is very fast, but after deployment is very slow.
I noticed a peak in the upload stream every time i change the dataset, often even higher than the download peak.
It seems like it's first downloading the dataset and than uploading it to the server. Fact is that the data is already server-side, loaded through a library.
My question is this, it's normal this peak in upload after changing the dataset? Because it feels very strange to me, and maybe i have simply done something wrong.
I noticed a little detail. Loading the fisrt dataset is followed by a long stream in upload, while loading the second has a shorter stream in upload.
I noticed that the load isn't actually slow. The fact is that, if i load the first dataset and wait for the upload peak to finish, then the second dataset will load immediately.
If instead i load the first dataset, and change to the second dataset without waiting the vanishing of the upload peak, it won't load the second dataset until the fisrt upload stream isn't finished.
So, it's not slow in loading the data, there's simply a concurrency of requests. The download of the second dataset is waiting to the first upload to be finished.
Still I don't understand what's happening, but maybe this information can help you to help me.
OK, i understood the problem. The upload stream i scaused by the client sending the list of visualized rows, plus the list of all possible row numbers. I don't know why this list isn't deduced from the dataset itself serverside, but the fact is that this list is 2.2Mb and with a slow connection in upload the time for the upload is considerable.
I will open a ticket to the DT page, and ask if this behavior is configurable or if it's possible to fix it.
