[
https://issues.apache.org/jira/browse/CASSANDRA-12915?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15693342#comment-15693342
]
Corentin Chary commented on CASSANDRA-12915:
--------------------------------------------
I also have no idea what the primary expression was supposed to be ... I guess
we can drop it.
It's true what we should do something before fetching the tokens on every
expressions. There are other things I was thinking about:
* Ordering expressions with a score (EQ first, then something like estimated
result rows)
* Skipping the following indexes if we already found one with less tokens than
command.limits().count()
Regarding 100000, that's the value at which I found iterating on the disk token
index was significantly slower than doing anything else. I guess using limits()
could work too. Regarding the factor, that's true, but I'm not sore what would
be a better heuristic.
In my case I do not use only EQ relations, in some cases these may be prefixes
(https://github.com/criteo/biggraphite/blob/9d436b6212b1f603fd3d4eff0b8c465d9a602454/biggraphite/drivers/cassandra.py,
https://github.com/criteo/biggraphite/blob/fcbb91bbad2149334d67082cc7f1c1573215c497/CASSANDRA_DESIGN.md)
So, what we need first is something like getEstimatedResultRows. Do you have a
good suggestion to do that without doing the search ? I didn't find much appart
from the index file size.
> SASI: Index intersection can be very inefficient
> ------------------------------------------------
>
> Key: CASSANDRA-12915
> URL: https://issues.apache.org/jira/browse/CASSANDRA-12915
> Project: Cassandra
> Issue Type: Improvement
> Components: sasi
> Reporter: Corentin Chary
> Fix For: 3.x
>
>
> It looks like RangeIntersectionIterator.java and be pretty inefficient in
> some cases. Let's take the following query:
> SELECT data FROM table WHERE index1 = 'foo' AND index2 = 'bar';
> In this case:
> * index1 = 'foo' will match 2 items
> * index2 = 'bar' will match ~300k items
> On my setup, the query will take ~1 sec, most of the time being spent in
> disk.TokenTree.getTokenAt().
> if I patch RangeIntersectionIterator so that it doesn't try to do the
> intersection (and effectively only use 'index1') the query will run in a few
> tenth of milliseconds.
> I see multiple solutions for that:
> * Add a static thresold to avoid the use of the index for the intersection
> when we know it will be slow. Probably when the range size factor is very
> small and the range size is big.
> * CASSANDRA-10765
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
