[
https://issues.apache.org/jira/browse/CASSANDRA-8731?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14305419#comment-14305419
]
Benedict commented on CASSANDRA-8731:
-------------------------------------
Well, the simplest implementation is very trivial: Instead of one PQ, you have
a PQ for each level of clustering column. Really not very advanced at all.
If you want to split intra-clustering-component to make byte-order comparable
fields more efficient, that is more involved.
That said, there is no harm in having both. It seems to me that it may be more
involved to do what you suggest, though, especially for sstables that _do_
overlap, but don't overlap for their entirety. At the moment we just lump all
our iterators together; slicing them and merging them independently actually
sounds to me to be much more fiddly and tough to get right than the simplest
approach outlined here.
> Optimise merges involving multiple clustering columns
> -----------------------------------------------------
>
> Key: CASSANDRA-8731
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8731
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Benedict
> Labels: performance
> Fix For: 3.0
>
>
> Since the new storage format is dead in the water for the moment, we should
> do our best to optimise current behaviour. When merging data from multiple
> sstables with multiple clustering columns, currently we must incur the full
> costs of comparison for the entire matching prefix, and must heapify every
> cell in our PriorityQueue, incurring lg(N) of these costlier comparisons for
> every cell we merge, where N is the number of sources we're merging.
> Essentially I'm proposing a trie-based merge approach as a replacement for
> the ManyToOne MergeIterator, wherein we treat each clustering component as a
> tree underwhich all Cells with a common prefix occur. We then perform a tree
> merge, rather than a flat merge. For byte-order fields this trie can even be
> a full binary-trie (although built on the fly). The advantage here is that we
> rapidly prune merges involving disjoint ranges, so that instead of always
> incurring lg(N) costs on each new record, we may often incur O(1) costs. For
> timeseries data, for instance, we could merge dozens of files and so long as
> they were non-overlapping our CPU burden would be little more than reading
> from a single file.
> On top of this, we no longer incur any of the shared prefix repetition costs,
> since we compare each prefix piece-wise, and only once.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
