[
https://issues.apache.org/jira/browse/LUCENE-2380?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12871342#action_12871342
]
Michael McCandless commented on LUCENE-2380:
--------------------------------------------
I did some rough estimates of RAM usage for StringIndex (trunk) vs
TermIndex (patch).
Java String is an object, so estimate 8 byte object header in the JRE.
It seems to have 3 int fields (offset, count, hashCode), from
OpenJDK's sources, plus ref to char[].
The char[] has 8 byte object header, int length, and actual array
data.
So in trunk's StringIndex:
per-unique-term: 40 bytes (48 on 64bit jre) + 2*length-of-string-in-UTF16
per-doc: 4 bytes (8 bytes on 64 bit)
In the patch:
per-unique-term: ceil(log2(totalUTF8BytesTermData)) + utf8 bytes + 1 or 2
bytes (vInt, for term length)
per-doc: ceil(log2(numUniqueTerm)) bits
So eg say you have an English title field, avg length 40 chars, and
assume always unique. On a 5M doc index, trunk would take ~591MB and
patch would take ~226 MB (32bit JRE) = 62% less.
But if you have a CJK title field, avg 10 chars (may be highish), it's
less savings because UTF8 takes 50% more RAM than UTF16 does for CJK
(and others). Trunk would take ~305MB and patch ~178MB (32bit JRE) =
42% less.
Also don't forget the GC load of having 5M String & char[] objects...
> Add FieldCache.getTermBytes, to load term data as byte[]
> --------------------------------------------------------
>
> Key: LUCENE-2380
> URL: https://issues.apache.org/jira/browse/LUCENE-2380
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Search
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: 4.0
>
> Attachments: LUCENE-2380.patch, LUCENE-2380.patch, LUCENE-2380.patch
>
>
> With flex, a term is now an opaque byte[] (typically, utf8 encoded unicode
> string, but not necessarily), so we need to push this up the search stack.
> FieldCache now has getStrings and getStringIndex; we need corresponding
> methods to load terms as native byte[], since in general they may not be
> representable as String. This should be quite a bit more RAM efficient too,
> for US ascii content since each character would then use 1 byte not 2.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
