https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6956

Mark Martinec <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #4 from Mark Martinec <[email protected]> ---
Works fine and fast, closing.

Btw, tested also with Redis 2.8-rc, no surprises there.
Just a redis server restart was needed, it picked the existing
database from its disk copy and resumed work.

Same speed with 2.8 as with 2.6.14 (or maybe a trifle faster
or am I imagining, no statistical significance - possibly due to
a fresh server start).

For fun I re-enabled our PostgreSQL bayes backend and compared timings.

The tok_get_all (i.e. lookup on all tokens) takes about 6 ms
on average with Redis+Lua, and SQL shows two timing clusters,
one around 10 ms, the other around 28 ms.

The tok_touch_all (i.e. updating atime / expiration time of
significant tokens, not autolearning) shows much higher difference:
redis+Lua takes about 1 ms, while an SQL update takes about 20 ms
for an average update.

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to