Open maxheld83 opened 4 years ago
nope it's not the data object, or at least not just them.
pryr::object_size(hoad::hybrid_publications(), hoad::unpaywall_df())
yields 56.9 MB which is not nothing, but doesn't eat up 4GiB of RAM.
Will need to investigate further, later.
now that we're on GCP (instead of shinyapps.io) as per #25, we're getting a clearer view into the resource utilisation of our shiny app.
It ain't pretty.
It (apparently, pending further testing) won't be stable below 4GiB of RAM.
I am guessing this is mostly the pretty huge data object being loaded into RAM, which is just R design.
Still, I should profile this to be sure that's what's going on.
Long-term, to make the app easier on the RAM, it would be good to offload most of this work to Google BigQuery or a separate R (plumber) API.
(We're not using much CPU, as expected).