I have a stable and fast internet connection. When I upgrade the system using pacman -Syu, it downloads the package files with higher size first and smallest last in descending order. So huge file size like MBs are downloaded with good speed but later on small size files in KB are downloaded very slow. The main issue is the number of such small size files in KB are in 100s and though my internet speed is good, it downloads very slowly as it has to fetch each file from the server and as a result, if I have 200 such files to be downloaded it takes at least 400-800 seconds which in total size could be around 50MBs which if was a single file could have been downloaded within 5-10 seconds.
Please once check below link for a video snippet for more clarity. Sorry I could not explain it clearly. Thanks https://drive.google.com/file/d/1THbIcU8DJlWut0kCcE_8Sw6Frw6ziSEn/view?usp=sharing
I don’t think it’s the problem of mirrors. Initially, it would download at around 5-7MB/s for large files. But for small KB-sized files, it downloads at 200KB/s. Can there be a speed difference for different files in the same mirror? Thanks
I don’t know. It’s possible, but if so, then this would be specific to that particular mirror. And in that case, it would be a good idea to switch mirrors.
Not a clear cut solution but it will definitely help
You can try enabling parallel downloading in pacman. To do that open /etc/pacman.conf as root and find and uncomment ParallelDownloads = 5.
Also (from my experience) make sure your /etc/pacman.d/mirrorlist doesn’t contain mirrors.piconets.webwerks.in/manjaro-mirror... mirror…it never works.
When retrieving a file, there is initially a little back and forth between the sender and receiver. At this point, latency plays a bigger role.
Later, when the download has started, the latency is no longer that important. The transmission speed is then important for this.
A parallel download will definitely improve this, because the handshake then plays less of a role at the beginning. You can use a number that roughly corresponds to the number of CPU cores (or cores*2 for hyperthreading). Opinions differ on cores +1 or cores -1 because everyone has different priorities.
