Closed trandersen-ufst closed 2 years ago
I am creating a pull request from my fork where I cannot create issues.. I was expecting to be able to create a properly named branch after creating this issue. Apparently this is not how it is done. I will submit a proper pull request in a few days.
"master" as of now takes 50 minutes on my Windows 10 machine. Initial experiments with appropriate buffering gives a run time of 2-3 minutes.
Got down to 44 seconds. Pull request at https://github.com/stevespringett/nist-data-mirror/pull/116
I noticed a performance issue too. Moving from one VM to another, the disk must be configured differently since running the mirror went from about 5 minutes to 45 minutes and seems to be mostly due to the lack of I/O buffering.
Thanks @trandersen-ufst for your PR. I came here to submit the same thing.
@sellersj You're welcome. Now a month later we ended up not using this but instead saving the dataDirectory of a normal dependency-check:check run in a SNAPSHOT Maven artifact which can then be unpacked when needed for "almost-offline" usage in our CICD build server. As the download+update bit is then decoupled from the actual check, we expect this to be very robust.
PR https://github.com/stevespringett/nist-data-mirror/pull/116 merged so I think that this is resolved. I tested it with the 1.6.0 release and it was much faster.
Fantastic! Thanks for testing.
Downloading and uncompressing from scratch takes almost 3 minutes on my machine. Initial profiling showed that InputStreams and OutputStreams are not buffered.