[Date Prev][Date Next]
[Chronological]
[Thread]
[Top]
Re: LMDB b-tree question
- To: Kristoffer Sjögren <stoffe@gmail.com>
- Subject: Re: LMDB b-tree question
- From: Howard Chu <hyc@symas.com>
- Date: Wed, 11 Feb 2015 20:13:50 +0000
- Cc: openldap-technical@openldap.org
- In-reply-to: <CADSBNAhEj7KVjbu=XT3KRBGQMoggqcf-fx12pT=p6SmaZKaKZQ@mail.gmail.com>
- References: <CADSBNAgHJPUwDzUaPoviWb=jST98hiVfMmaW4cgymnJZnrnx5A@mail.gmail.com> <54DB93A7.6070102@symas.com> <CADSBNAhEj7KVjbu=XT3KRBGQMoggqcf-fx12pT=p6SmaZKaKZQ@mail.gmail.com>
- User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:37.0) Gecko/20100101 Firefox/37.0 SeaMonkey/2.34a1
Kristoffer Sjögren wrote:
Thanks Howard. It wasn't immediately evident in the wording
that MDB_stat use an exclusive b-tree.
Is there some way of estimating the depth of the b-tree based on
database size, number of entries or other metrics?
????
mdb_stat already reports the depth of the tree.
On Wed, Feb 11, 2015 at 6:38 PM, Howard Chu <hyc@symas.com
<mailto:hyc@symas.com>> wrote:
Kristoffer Sjögren wrote:
Hi
Does every database consist of a separate b-tree or is
"database" just a
name spacing convince? I.e. are reads faster for many smaller
databases
rather than one large database?
As the documentation says, named databases are separate b-trees.
Reads may be slightly faster, sure, but not by much. You need to
have enough separate DBs to reduce the overall tree height of the
equivalent combined tree to see a measurable difference.
--
-- Howard Chu
CTO, Symas Corp. http://www.symas.com
Director, Highland Sun http://highlandsun.com/hyc/
Chief Architect, OpenLDAP http://www.openldap.org/project/