Skip to content

Comment on Save Redis memory with large list compressionparent

Comments

The blog post you referenced I think was done by instagram, not tumblr. I remember reading it and doing something similar. They also found doubling the `hash-max-ziplist-entries` options to 512 from 1024 also didn't cause much CPU load.

Something you can also try as well is encoding with msgpack rather than JSON. You can save 30% from the original JSON string without losing the keynames.

Indeed it was Instagram, I stand corrected.

I'm aware of Msgpack and I was under the impression that it doesn't handle UTF data very well, however that was some time ago and I may be mis-remembering. Msgpack libraries are also supposed to be quite a bit faster at encoding/decoding, right?

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.