Hacker News new | ask | show | jobs
by chrismorgan 1600 days ago
2–3 seconds for 75 pages is not at all fast. My own Zola-powered site with unusually complex templates and emitting a bunch of large feeds takes under 200ms for that many pages; and Zola is not really architected to be fast (it’s more just incidentally comparatively fast because of Rust)—a static site generator actually designed for performance could certainly handle a site like that in well under fifty milliseconds with no loss of ergonomics, and I think five milliseconds with potential slight loss of ergonomics is feasible.

(Amusing note: Hexo and Zola both have a heading on their home pages “blazing fast”. I think they’re both wrong.)
1 comments
shakna 1600 days ago
Once you get below a couple seconds, you're unlikely to be measuring the speed of the SSG, you're measuring I/O latency. Suggesting it could be faster or slower is pointless unless you have identical hardware. The same disks, CPU, and RAM, the same OS, down to the kernel version. The same running processes.
link
dmitriid 1600 days ago
Even a 5400 rpm hard disk delivers reads in the hundreds of megabytes per second.

75 pages should never take several seconds. Or even "under a couple of seconds". It should bot even register in the radar.

link
chrismorgan 1600 days ago
Hundreds of megabytes per second is decidedly on the high side for such a device to begin with, but it only holds for big sequential writes. For random I/O, which is what this is, you’re more likely to average 0.5–5MB/s plus latency.

I was assuming it would be more CPU- than I/O-bound, because hopefully we’re all using SSDs these days, and because Hexo’s website uses similar time figures in its preposterous “blazing fast” description.

shakna’s gone too far the other way (a couple of seconds for a load like this (~80 file reads, processing, ~80 writes, all of the files under 64KB) is easily enough that if it’s measuring I/O latency there’s either something wrong with your machine or you have a slow HDD and wrote your I/O code the wrong way), but my 200ms certainly already reaches the point where CPU, disk and OS/antivirus solution all matter significantly.

link