I saw that the ratio of reads to inserts in our hasherv is ~10:1. So I have added a cache for unihash reads to somewhat off-load the hashserver. This since it rapidly becomes CPU bound and also that it is unresponsive during cleanup operations like vacuum. This goes for the classic sqlite based solution.
We have run this cache in production now with and it seems to do OK so I would like to share it either as a contribution or an idea for improvement. Tobias Hagelborn (1): hashserv: Unihash cache lib/hashserv/server.py | 188 ++++++++++++++++++++++++++++++++++------- 1 file changed, 159 insertions(+), 29 deletions(-) -- 2.30.2
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#191479): https://lists.openembedded.org/g/openembedded-core/message/191479 Mute This Topic: https://lists.openembedded.org/mt/102890272/21656 Group Owner: openembedded-core+ow...@lists.openembedded.org Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
