From: | Leonardo Francalanci <m_lists(at)yahoo(dot)it> |
---|---|
To: | Leonardo Francalanci <m_lists(at)yahoo(dot)it>, Merlin Moncure <mmoncure(at)gmail(dot)com> |
Cc: | pgsql-hackers <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Fast insertion indexes: why no developments |
Date: | 2013-10-29 15:49:43 |
Message-ID: | 1383061783.2243.YahooMailNeo@web172602.mail.ir2.yahoo.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
> Another point to add: I don't really see btree as a barrier to
> performance for most of the problems I face. The real barriers to
> database performance are storage, contention, and query planning.
Ehm that's true for regular OLTP stuff, which I understand is what most (95%?) of people use/need. But if you try to insert rows into a 50M table with a couple of indexes, btrees just can't keep up.
Of course, you can't have it all: fast at big table insertion, good contention, good query times...
> Postgres btreee indexes are pretty fast and for stuff like bulk
> insertions there are some optimization techniques available (such as
> sharding or create index concurrently).
At the moment I'm relying on partitioning + creating indexes in bulk on "latest" table (the partitioning is based on time). But that means K*log(N) search times (where K is the number of partitions).
That's why I gave a look at these different indexing mechanisms.
From | Date | Subject | |
---|---|---|---|
Next Message | Leonardo Francalanci | 2013-10-29 15:53:40 | Re: Fast insertion indexes: why no developments |
Previous Message | Andres Freund | 2013-10-29 15:43:26 | Re: logical changeset generation v6.2 |