accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Keith Turner (JIRA)" <>
Subject [jira] [Commented] (ACCUMULO-4468), PartialKey) improvement
Date Thu, 22 Sep 2016 15:28:22 GMT


Keith Turner commented on ACCUMULO-4468:

bq. That said, maybe there would be benefits to storing all the pieces of the Key in a single
byte array, and maintaining indices into it to track the individual parts, rather than several
smaller arrays...

Key used to be like that (one large byte array with pointers).  I changed it when adding support
for relative compression in rfile.  The reasoning behind the change was so that when rfile
deserializes a key and a field is the same as the last key, it can just point to the previous
byte array.  This makes equality comparisons on rows or columns that are the same really fast
(because the byte array is the same and equality checks that).  The code that serializes keys
and transfers them across the network also does this.  So it may be interesting to have the
test stream keys from an RFile.

bq. I think it would be worth revisiting the comparison mechanism in isEqual, too, doing something
like the Unsafe method used in Hadoop's FastByteComparisons class but going in reverse.

Thats sounds like an interesting line of investigation.  The compare methods could also leverage
this technique.  We may have an issue open for this already.

>, PartialKey) improvement
> ----------------------------------------------------------
>                 Key: ACCUMULO-4468
>                 URL:
>             Project: Accumulo
>          Issue Type: Improvement
>          Components: core
>    Affects Versions: 1.8.0
>            Reporter: Will Murnane
>            Priority: Trivial
>              Labels: newbie, performance
>         Attachments: benchmark.tar.gz, key_comparison.patch
> In the Key.equals(Key, PartialKey) overload, the current method compares starting at
the beginning of the key, and works its way toward the end. This functions correctly, of course,
but one of the typical uses of this method is to compare adjacent rows to break them into
larger chunks. For example, accumulo.core.iterators.Combiner repeatedly calls this method
with subsequent pairs of keys.
> I have a patch which reverses the comparison order. That is, if the method is called
with ROW_COLFAM_COLQUAL_COLVIS, it will compare visibility, cq, cf, and finally row. This
(marginally) improves the speed of comparisons in the relatively common case where only the
last part is changing, with less complex code.

This message was sent by Atlassian JIRA

View raw message