Commit Graph

7 Commits

Author SHA1 Message Date
Tomislav Novak
35c8c814e8 Make ForwardIterator::status() more efficient
Summary:
In D19581 I made `ForwardIterator::status()` check all child iterators,
including immutable ones. It's, however, not necessary to do it every
time -- it'll suffice to check only when they're used and their status
could change.

This diff:
* introduces `immutable_status_` which is updated by `Seek()` and `Next()`
* removes special handling of `kIncomplete` status in those methods

Test Plan:
* `db_test`
* hacked ReadSequential in db_bench.cc to check `status()` in addition to
  validity:

```
   $ ./db_bench -use_existing_db -benchmarks readseq -disable_auto_compactions \
      -use_tailing_iterator  # without this patch
   Keys:       16 bytes each
   Values:     100 bytes each (50 bytes after compression)
   Entries:    1000000
   [...]
   DB path: [/dev/shm/rocksdbtest/dbbench]
   readseq      :       0.562 micros/op 1778103 ops/sec;   98.4 MB/s
   $ ./db_bench -use_existing_db -benchmarks readseq -disable_auto_compactions \
      -use_tailing_iterator  # with the patch
   readseq      :       0.433 micros/op 2311363 ops/sec;  127.8 MB/s
```

Reviewers: igor, sdong, ljin

Reviewed By: ljin

Subscribers: dhruba, leveldb, march, lovro

Differential Revision: https://reviews.facebook.net/D24063
2014-11-10 15:44:20 -08:00
Lei Jin
574028679b dynamic max_sequential_skip_in_iterations
Summary:
This is not a critical options. Making it dynamic so that we can remove
more reference to cfd->options()

Test Plan: unit test

Reviewers: yhchiang, sdong, igor

Reviewed By: igor

Subscribers: leveldb

Differential Revision: https://reviews.facebook.net/D24957
2014-10-23 15:34:21 -07:00
Tomislav Novak
187b29938c ForwardIterator: update prev_key_ only if prefix hasn't changed
Summary:
Since ForwardIterator is on a level below DBIter, the latter may call Next() on
it (e.g. in order to skip deletion markers). Since this also updates
`prev_key_`, it may prevent the Seek() optimization.

For example, assume that there's only one SST file and it contains the following
entries: 0101, 0201 (`ValueType::kTypeDeletion`, i.e. a tombstone record), 0201
(`kTypeValue`), 0202. Memtable is empty. `Seek(0102)` will result in `prev_key_`
being set to `0201` instead of `0102`, since `DBIter::Seek()` will call
`ForwardIterator::Next()` to skip record 0201. Therefore, when `Seek(0102)` is
called again, `NeedToSeekImmutable()` will return true.

This fix relies on `prefix_extractor_` to detect prefix changes. `prev_key_` is
only set to `current_->key()` as long as they have the same prefix.

I also made a small change to `NeedToSeekImmutable()` so it no longer returns
true when the db is empty (i.e. there's nothing but a memtable).

Test Plan:
   $ TEST_TMPDIR=/dev/shm/rocksdbtest ROCKSDB_TESTS=TailingIterator ./db_test

Reviewers: sdong, igor, ljin

Reviewed By: ljin

Subscribers: leveldb

Differential Revision: https://reviews.facebook.net/D23823
2014-10-01 17:10:48 -07:00
Stanislau Hlebik
45a5e3ede0 Remove path with arena==nullptr from NewInternalIterator
Summary:
Simply code by removing code path which does not use Arena
from NewInternalIterator

Test Plan:
make all check
make valgrind_check

Reviewers: sdong

Reviewed By: sdong

Subscribers: leveldb

Differential Revision: https://reviews.facebook.net/D22395
2014-09-04 17:40:41 -07:00
Tomislav Novak
0f9c43ea36 ForwardIterator: reset incomplete iterators on Seek()
Summary:
When reading from kBlockCacheTier, ForwardIterator's internal child iterators
may end up in the incomplete state (read was unable to complete without doing
disk I/O). `ForwardIterator::status()` will correctly report that; however, the
iterator may be stuck in that state until all sub-iterators are rebuilt:

  * `NeedToSeekImmutable()` may return false even if some sub-iterators are
    incomplete
  * one of the child iterators may be an empty iterator without any state other
    that the kIncomplete status (created using `NewErrorIterator()`); seeking on
    any such iterator has no effect -- we need to construct it again

Akin to rebuilding iterators after a superversion bump, this diff makes forward
iterator reset all incomplete child iterators when `Seek()` or `Next()` are
called.

Test Plan: TEST_TMPDIR=/dev/shm/rocksdbtest ROCKSDB_TESTS=TailingIterator ./db_test

Reviewers: igor, sdong, ljin

Reviewed By: ljin

Subscribers: lovro, march, leveldb

Differential Revision: https://reviews.facebook.net/D22575
2014-08-29 16:21:29 -07:00
Igor Canadi
fd27001072 Fix compile errors on Mac
Summary: https://phabricator.fb.com/P11372644

Test Plan: compiles

Reviewers: sdong, ljin

Reviewed By: ljin

Subscribers: leveldb

Differential Revision: https://reviews.facebook.net/D18873
2014-06-03 12:28:58 -07:00
Lei Jin
388d2054c7 forward iterator
Summary:
Forward iterator puts everything together in a flat structure instead of
a hierarchy of nested iterators. this should simplify the code and
provide better performance. It also enables more optimization since all
information are accessiable in one place.
Init evaluation shows about 6% improvement

Test Plan: db_test and db_bench

Reviewers: dhruba, igor, tnovak, sdong, haobo

Reviewed By: haobo

Subscribers: sdong, leveldb

Differential Revision: https://reviews.facebook.net/D18795
2014-05-30 14:31:55 -07:00