Whilst download size is larger, it is at no additional cost to our end users, especially in the cloud. Majority of our clouds offer free transfers inside the availability zone for which mirrors are provided. Obviously it will increase the disk/cache storage size requirements of our mirrors.
Our end users have about 3 sets of kernels installed on disk, for which they do pay more per GB between 2 to 12 cents (depending on technology IOPS / hdd / sda / nvme). All things being equal our users would choose larger download sizes to gain smaller install size. I assume that we have vastly larger install base, compared to number of our mirrors. I also hope that our install base is not larger than the total bandwidth available of our mirrors, such that this switch would cause collapse of our mirror network. I also notice that kernel uses default options for zstd module compression, instead of using -19 like it recommends for the initrds. I also notice that kmod in focal doesn't support zstd. Imho we should fix above before turning this on. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1932329 Title: Benchmark if we can compress kernel modules To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1932329/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
