Performance drop when adding packets to a remote feed

[Oli_Wachno]

Hi there,

We are seeing a bit of a performance (though I wouldn't exactly call it) issue I'd like to understand.

We have a rather large repo on our network which holds all our packets for a LabVIEW installation.

Since we don't want to install from the network, we are downloading the stuff into a local repo first.

nipkg.exe feed-create \local_temp \\unc_path_to_remote_repo 
# feed-add takes aaaaaaaaages
nipkg.exe feed-add -n="myfeed" \local_temp
#download actually pretty quick
nipkg.exe feed-download "myfeed" \local_feed_temp
nipkg.exe feed-remove myfeed
nipkg.exe feed-create \local_feed \local_feed_temp 
# adding from the local location works in a wink of an eye
nipkg.exe feed-add -n="localfeed" \local_feed

I am aware, that there is a lot of directory shuffeling going on, please ignore. We had to do this for some reason.
Any idea, how we can optimise the first feed-add?

Cheers

Oli

[CL_eisenwr]

I have also seen this. It seems like NIPM is always rewriting the whole file package by package so there is a lot of network traffic instead of just creating a file locally in memory/disk and copying it out. We were adding a single package to a feed and it would take ~5 minutes for the action to complete. I had mentioned this to NI years/several owners ago, but I am not sure if anything was ever done about it.

[Scott_Richardson]

Oli -

You are creating a feed with references to packages on the network to then use NIPM to download the packages. Is there a reason that you could not just copy the *.nipkg files from the network to the local directory and not use NIPM?

I do not remember, but it might be that NIPM is validating the package in some way in addition to downloading it and that is the unwanted overhead. I believe that the same validation will occur when you later create the feed again locally.

So I am suggesting, (a) copy packages locally, (b) create feed in directory where the packages are located, (c) register feed locally.