Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.
Sign upSuboptimal block sizes #3
Comments
This comment has been minimized.
This comment has been minimized.
|
Yea, good point! I picked 8kb as its a nice default for packet size in networking. I think there's perhaps two things we could do, maybe one or the other, or both:
Perhaps the best is to have the config option, and prefer that over |
This comment has been minimized.
This comment has been minimized.
|
Hmm, wasn't this closed in #5? |
lnicola
closed this
Dec 20, 2017
lnicola
referenced this issue
Mar 17, 2018
Closed
Use the fs block size for chunks if larger than 64 kb #3
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
lnicola commentedNov 14, 2017
•
edited
While
8192 bseems like a reasonable default for reading files, on modern systems it might be a bit too small, especially for sequential access. We should consider bumping it up a bit.I ran a couple of tests on my NAS:
Note that I'm using ZFS, which doesn't support
sendfile(). The128 kbvalue is the filesystem preferred buffer size (st_blksizeinstruct stat).EDIT: notice how👍 .
hyperwithfutures-fswas faster thannginx