|
|
|
|
|
|
by dmw_ng
1614 days ago
|
|
|
You can squeeze a hell of a lot more perf out of general purpose compute that is maybe an extra 15 ms away than one-shot workers lacking any in-RAM persistence, or ability to do much more than limited local k/v lookups. The equation is totally different. Many workers I've seen (or written) start by reaching out over the Internet to another backend. There goes most of your edge benefit almost immediately. |
|
|