Summary:
MyRocks calls `GetForUpdate` on `INSERT`, for unique key check, and in almost all cases GetForUpdate returns empty result. For such cases, whole key bloom filter is helpful.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4985
Differential Revision: D14118257
Pulled By: miasantreble
fbshipit-source-id: d35cb7109c62fd5ad541a26968e3a3e16d3e85ea
Summary:
The info log header feature never worked well, because log level Header was not
translated to Logger::LogHeader() call. Fix it.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4980
Differential Revision: D14087283
Pulled By: siying
fbshipit-source-id: 7e7d03ce35fa8d13d4ee549f46f7326f7bc0006d
Summary:
Right now when a flush is triggered, the memory consumption is logged but data size is not.
It's useful to log both when we debug unexpected small flushed file size.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4979
Differential Revision: D14071979
Pulled By: siying
fbshipit-source-id: 0cd60449c5205eb00e0fbc299084418f609904ed
Summary:
We introduced ttl option in CompactionOptionsFIFO when ttl-based file
deletion (compaction) was supported only as part of FIFO Compaction. But
with the extension of ttl semantics even to Level compaction,
CompactionOptionsFIFO.ttl can now be deprecated. Instead we will start
using ColumnFamilyOptions.ttl for FIFO compaction as well.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4965
Differential Revision: D14072960
Pulled By: sagar0
fbshipit-source-id: c98cc2ae695a28136295787cd88d36a220fc219e
Summary:
If `CompressionOptions::max_dict_bytes` and/or `CompressionOptions::zstd_max_train_bytes` are set, `SstFileWriter` will now generate files respecting those options.
I refactored the logic a bit for deciding when to use dictionary compression. Previously we plumbed `is_bottommost_level` down to the table builder and used that. However it was kind of confusing in `SstFileWriter`'s context since we don't know what level the file will be ingested to. Instead, now the higher-level callers (e.g., flush, compaction, file writer) are responsible for building the right `CompressionOptions` to give the table builder.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4978
Differential Revision: D14060763
Pulled By: ajkr
fbshipit-source-id: dc802c327896df2b319dc162d6acc82b9cdb452a
Summary:
if an operation just involves a single column family, then we do
not have to set the kInAtomicGroup tag when writing to MANIFEST. This change
can fix a compatibility test failure, i.e. 5.15 and earlier cannot recognize
kInAtomicGroup tag.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4981
Differential Revision: D14072687
Pulled By: riversand963
fbshipit-source-id: 46b0c61e399f16c6b7169de0b33430d0ed90d6d4
Summary:
Make file ingestion atomic.
as title.
Ingesting external SST files into multiple column families should be atomic. If
a crash occurs and db reopens, either all column families have successfully
ingested the files before the crash, or non of the ingestions have any effect
on the state of the db.
Also add unit tests for atomic ingestion.
Note that the unit test here does not cover the case of incomplete atomic group
in the MANIFEST, which is covered in VersionSetTest already.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4895
Differential Revision: D13718245
Pulled By: riversand963
fbshipit-source-id: 7df97cc483af73ad44dd6993008f99b083852198
Summary:
Previously, stats were logged in warning level. This was done in that way because
people reported that it wasn't logged in MyRocks. However, later we learned that it turns
out to be due to a bug in MyRocks, which is fixed in
79bb705e74
Now we revert the stats logging to INFO level, so that it doesn't pollute the warning
level logging.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4977
Differential Revision: D14058485
Pulled By: siying
fbshipit-source-id: 19fab323c19d9bc88184287f209551f9a77ca0e6
Summary:
In `DBImpl::AtomicFlushMemTablesToOutputFiles`, we need to call fsync only once
on the same data directory. If two column families share a common directory for
their data, we call fsync only once.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4817
Differential Revision: D13543689
Pulled By: riversand963
fbshipit-source-id: 4701d77c96a47802fbf6cb9f3337ee65d46b95f5
Summary:
Our previous approach was to train one compression dictionary per compaction, using the first output SST to train a dictionary, and then applying it on subsequent SSTs in the same compaction. While this was great for minimizing CPU/memory/I/O overhead, it did not achieve good compression ratios in practice. In our most promising potential use case, moderate reductions in a dictionary's scope make a major difference on compression ratio.
So, this PR changes compression dictionary to be scoped per-SST. It accepts the tradeoff during table building to use more memory and CPU. Important changes include:
- The `BlockBasedTableBuilder` has a new state when dictionary compression is in-use: `kBuffered`. In that state it accumulates uncompressed data in-memory whenever `Add` is called.
- After accumulating target file size bytes or calling `BlockBasedTableBuilder::Finish`, a `BlockBasedTableBuilder` moves to the `kUnbuffered` state. The transition (`EnterUnbuffered()`) involves sampling the buffered data, training a dictionary, and compressing/writing out all buffered data. In the `kUnbuffered` state, a `BlockBasedTableBuilder` behaves the same as before -- blocks are compressed/written out as soon as they fill up.
- Samples are now whole uncompressed data blocks, except the final sample may be a partial data block so we don't breach the user's configured `max_dict_bytes` or `zstd_max_train_bytes`. The dictionary trainer is supposed to work better when we pass it real units of compression. Previously we were passing 64-byte KV samples which was not realistic.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4952
Differential Revision: D13967980
Pulled By: ajkr
fbshipit-source-id: 82bea6f7537e1529c7a1a4cdee84585f5949300f
Summary:
If IsInSnapshot(seq2, snapshot) determines that the snapshot is released, the future queries IsInSnapshot(seq1, snapshot) could still return a definitive answer of true if for example seq1 is too old that is determined visible in all snapshots. This violates a recently added assert statement to compaction iterator. The patch relaxes the assert.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4969
Differential Revision: D14030998
Pulled By: maysamyabandeh
fbshipit-source-id: 6db53db0e37d0a20e8997ef2c1004b8627614ab9
Summary:
Always enable properties block checksum verification for block-based table. For external SST file ingested with 'write_global_seqno==true', we use 'DecodeEntrySlow' to parse its blocks' contents so that the process will not die upon failing the assertion possibly caused by corruption.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4956
Differential Revision: D14012741
Pulled By: riversand963
fbshipit-source-id: 8b766e6f54b36f8f9e074c0e19e0926ec3cce186
Summary:
Implement trace sampling to allow user to specify the sampling frequency, i.e. save one per how many requests, so that a user does not need to log all if he/she is interested in only a sampled set.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4963
Differential Revision: D14011190
Pulled By: tang-jianfeng
fbshipit-source-id: 078b631d9319b67cb089dd2c30e21d0df8dc406a
Summary:
We want to reserve some right that some extra information added manifest
in the future can be forward compatible by previous versions. Now we create a
place holder for that. A bit in tag is added to indicate that a field can be
safely ignored.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4960
Differential Revision: D14000484
Pulled By: siying
fbshipit-source-id: cbf5bad3f9d5ec798f789806f244d1c20d3b66d6
Summary:
We found that the behavior of CompactionFilter::IgnoreSnapshots() = false isn't
what we have expected. We thought that snapshot will always be preserved.
However, we just realized that, if no snapshot is created while compaction
starts, and a snapshot is created after that, the data seen from the snapshot
can successfully be dropped by the compaction. This creates a strange behavior
to the feature, which is hard to explain. Like what is documented in code
comment, this feature is not very useful with snapshot anyway. The decision
is to deprecate the feature.
We keep the function to avoid to break users code. However, we will fail
compactions if false is returned.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4954
Differential Revision: D13981900
Pulled By: siying
fbshipit-source-id: 2db8c2c3865acd86a28dca625945d1481b1d1e36
Summary:
Cuckoo Hash is less useful than we initially expected. Remove it.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4953
Differential Revision: D13979264
Pulled By: siying
fbshipit-source-id: 2a60afdaa989f045357398b43a1cc5d46f4492ed
Summary:
This will fix the following build error:
> db/db_test.cc: In member function ‘virtual void rocksdb::DBTest_CompactFilesShouldTriggerAutoCompaction_Test::TestBody()’:
> db/db_test.cc:5462:8: error: ‘class rocksdb::DB’ has no member named ‘GetColumnFamilyMetaData’
> db_->GetColumnFamilyMetaData(db_->DefaultColumnFamily(), &cf_meta_data);
> db/db_test.cc:5490:8: error: ‘class rocksdb::DB’ has no member named ‘GetColumnFamilyMetaData’
> db_->GetColumnFamilyMetaData(db_->DefaultColumnFamily(), &cf_meta_data);
> db/db_test.cc:5499:8: error: ‘class rocksdb::DB’ has no member named ‘GetColumnFamilyMetaData’
> db_->GetColumnFamilyMetaData(db_->DefaultColumnFamily(), &cf_meta_data);
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4950
Differential Revision: D13965378
Pulled By: miasantreble
fbshipit-source-id: a975435476fe555b1cd9d5da263ee3da3acdea56
Summary:
Existing implementation of PerfContext does not define copy constructor or assignment operator, which could potentially cause problems when user create copies and resets the builtin one. This PR address the issue by providing these two constructors with deep copy semantics.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4919
Differential Revision: D13960406
Pulled By: miasantreble
fbshipit-source-id: 36aab5aaee65d4480f537e4e22148faa45e8e334
Summary:
CompactFiles() may block auto compaction which could cuase DB hang when it
reachs level0_stop_writes_trigger.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4940
Differential Revision: D13929648
Pulled By: cooldoger
fbshipit-source-id: 10842df38df3bebf862cd1a120a88ce961fdd381
Summary:
In NotFound cases, stats BYTES_READ and perf_context.get_read_bytes is still be increased. The amount increased will be
whatever size of the string or PinnableSlice that users passed in as the output data structure. This is wrong. Fix this by not
increasing these two counters.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4938
Differential Revision: D13908963
Pulled By: siying
fbshipit-source-id: 60bce42e4fbb9862bba3da36dbc27b2963ea6162
Summary:
Fix the ouput overlap bug when using subcompactions, the upper bound of output
file was extended incorrectly.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4898
Differential Revision: D13736107
Pulled By: ajkr
fbshipit-source-id: 21dca09f81d5f07bf2766bf566f9b50dcab7d8e3
Summary:
The patch fixes the following analyze error by checking the return status of ParseInternalKey.
```
db/merge_helper.cc:306:23: warning: The right operand of '==' is a garbage value
assert(kTypeMerge == orig_ikey.type);
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4937
Differential Revision: D13908506
Pulled By: maysamyabandeh
fbshipit-source-id: 68d7771e75519da3d4bd807fd231675ec12093f6
Summary:
1. this commit fixes our handling of a combination of two separate edge
cases. If a flush job does not pick any memtable to flush (because another
flush job has already picked the same memtables), and the column family
assigned to the flush job is dropped right before RocksDB calls
rocksdb::InstallMemtableAtomicFlushResults, our original code passes
a FileMetaData object whose file number is 0, failing the assertion in
rocksdb::InstallMemtableAtomicFlushResults (assert(m->GetFileNumber() > 0)).
2. Also piggyback a small change: since we already create a local copy of column family's mutable CF options to eliminate potential race condition with `SetOptions` call, we might as well use the local copy in other function calls in the same scope.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4932
Differential Revision: D13901322
Pulled By: riversand963
fbshipit-source-id: b936580af7c127ea0c6c19ea10cd5fcede9fb0f9
Summary:
FlushMemTablesToOutputFiles calls FlushMemTableToOutputFile for each column family. The patch moves the take-snapshot logic to outside FlushMemTableToOutputFile so that it does it once for all the flushes. This also addresses a deadlock issue for resetting the managed snapshot of job_snapshot in the 2nd call to FlushMemTableToOutputFile.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4934
Differential Revision: D13900747
Pulled By: maysamyabandeh
fbshipit-source-id: f3cd650c5fff24cf95c1aaf8a10c149d42bf042c
Summary:
Measure CPU time consumed for a compaction and report it in the stats report
Enable NowCPUNanos() to work for MacOS
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4889
Differential Revision: D13701276
Pulled By: zinoale
fbshipit-source-id: 5024e5bbccd4dd10fd90d947870237f436445055
Summary:
before file ingestion (in preparation phase), verify the checksums of
the blocks of the external SST file, including properties block with global
seqno.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4916
Differential Revision: D13863501
Pulled By: riversand963
fbshipit-source-id: dc54697f970e3807832e2460f7228fcc7efe81ee
Summary:
While stepping through the code I noticed that there is a redundant call to TableFileName.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4925
Differential Revision: D13845749
Pulled By: sagar0
fbshipit-source-id: 31db45716b4d720e0e0350dd457b49d6f1848e7d
Summary:
Store_index_in_file is a less useful feature. To simplify the code to maintain, we are dropping the feature.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4914
Differential Revision: D13791883
Pulled By: siying
fbshipit-source-id: d187c5d662584866103e4b77d09dfb925509ae2e
Summary:
https://github.com/facebook/rocksdb/pull/4053 avoids memcopy for Get() results if files are immortable
(read-only DB, max_open_files=-1) and the file is ammaped. The same optimization is being applied to PlainTable
here.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4924
Differential Revision: D13827749
Pulled By: siying
fbshipit-source-id: 1f2cbfc530b40ce08ccd53f95f6e78de4d1c2f96
Summary:
I didn't find where customized hash function is used in DynamicBloom. This can only reduce performance. Remove it.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4915
Differential Revision: D13794452
Pulled By: siying
fbshipit-source-id: e38669b11e01444d2d782da11c7decabbd851819
Summary:
Previously compaction was not collapsing operands for a first
key on a layer, even in cases when it was its root of history. Some
tests (CompactionJobTest.NonAssocMerge) was actually accounting
for that bug,
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4909
Differential Revision: D13781169
Pulled By: finik
fbshipit-source-id: d2de353ecf05bec39b942cd8d5b97a8dc445f336
Summary:
- If block cache disabled or not used for meta-blocks, `BlockBasedTableReader::Rep::uncompression_dict` owns the `UncompressionDict`. It is preloaded during `PrefetchIndexAndFilterBlocks`.
- If block cache is enabled and used for meta-blocks, block cache owns the `UncompressionDict`, which holds dictionary and digested dictionary when needed. It is never prefetched though there is a TODO for this in the code. The cache key is simply the compression dictionary block handle.
- New stats for compression dictionary accesses in block cache: "BLOCK_CACHE_COMPRESSION_DICT_*" and "compression_dict_block_read_count"
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4881
Differential Revision: D13663801
Pulled By: ajkr
fbshipit-source-id: bdcc54044e180855cdcc57639b493b0e016c9a3f
Summary:
Even one key falls in a file's range, we can not infer it definitely exists in this file.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4902
Differential Revision: D13795018
Pulled By: siying
fbshipit-source-id: 590956f727e9440fcdee55ad9541ace934c64914
Summary:
compaction_pri = kMinOverlappingRatio usually provides much better write amplification than the default.
https://github.com/facebook/rocksdb/pull/4907 fixes one shortcome of this option. Make it default.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4911
Differential Revision: D13789262
Pulled By: siying
fbshipit-source-id: d90acf8c4dede44f00d183ca4c7a210259378269
Summary:
Right now, CompactionPri = kMinOverlappingRatio provides best write amplification, but it doesn't
prioritize files with more tombstones. We combine the two good features: make kMinOverlappingRatio
to boost files with lots of tombstones too.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4907
Differential Revision: D13788774
Pulled By: siying
fbshipit-source-id: 1991cbb495fb76c8b529de69896e38d81ed9d9b3
Summary:
This is essentially a re-submission of #4251 with a few improvements:
- Split `CompressionDict` into two separate classes: `CompressionDict` and `UncompressionDict`
- Eliminated `Init` functions. Instead do all initialization work in constructors.
- Added test case for parallel DB open, which is the scenario where #4251 failed under TSAN.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4849
Differential Revision: D13606039
Pulled By: ajkr
fbshipit-source-id: 08c236059798c710db9cbf545fce0f371232d447
Summary:
Fix how CompactionIterator::findEarliestVisibleSnapshots handles released snapshot. It fixing the two scenarios:
Scenario 1:
key1 has two values v1 and v2. There're two snapshots s1 and s2 taken after v1 and v2 are committed. Right after compaction output v2, s1 is released. Now findEarliestVisibleSnapshot may see s1 being released, and return the next snapshot, which is s2. That's larger than v2's earliest visible snapshot, which was s1.
The fix: the only place we check against last snapshot and current key snapshot is when we decide whether to compact out a value if it is hidden by a later value. In the check if we see current snapshot is even larger than last snapshot, we know last snapshot is released, and we are safe to compact out current key.
Scenario 2:
key1 has two values v1 and v2. there are two snapshots s1 and s2 taken after v1 and v2 are committed. During compaction before we process the key, s1 is released. When compaction process v2, snapshot checker may return kSnapshotReleased, and the earliest visible snapshot for v2 become s2. When compaction process v1, snapshot checker may return kIsInSnapshot (for WritePrepared transaction, it could be because v1 is still in commit cache). The result will become inconsistent here.
The fix: remember the set of released snapshots ever reported by snapshot checker, and ignore them when finding result for findEarliestVisibleSnapshot.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4890
Differential Revision: D13705538
Pulled By: maysamyabandeh
fbshipit-source-id: e577f0d9ee1ff5a6035f26859e56902ecc85a5a4
Summary:
Compaction iterator keep a copy of list of live snapshots at the beginning of compaction, and then query snapshot checker to verify if values of a sequence number is visible to these snapshots. However when the snapshot is released in the middle of compaction, the snapshot checker implementation (i.e. WritePreparedSnapshotChecker) may remove info with the snapshot and may report incorrect result, which lead to values being compacted out when it shouldn't. This patch conservatively keep the values if snapshot checker determines that the snapshots is released.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4858
Differential Revision: D13617146
Pulled By: maysamyabandeh
fbshipit-source-id: cf18a94f6f61a94bcff73c280f117b224af5fbc3
Summary:
By convention, time_t almost always stores the integral number of seconds since
00:00 hours, Jan 1, 1970 UTC, according to http://www.cplusplus.com/reference/ctime/time_t/.
We surely want more precision than seconds.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4868
Differential Revision: D13633046
Pulled By: riversand963
fbshipit-source-id: 4e01e23a22e8838023c51a91247a286dbf3a5396
Summary:
With WritePrepared transaction, flush/compaction can contain uncommitted keys, and those keys can get committed during compaction. If a snapshot is taken before the key is committed, it should not see the key. On the other hand, compaction grab the list of snapshots at its beginning, and only consider those snapshots to dedup keys. Consider the case:
```
seq = 1: put "foo" = "bar"
seq = 2: transaction T: delete "foo", prepare
seq = 3: compaction start
seq = 4: take snapshot S
seq = 5: transaction T: commit.
...
seq = N: compaction iterator reached key "foo".
```
When compaction start, the list of snapshot is empty. Compaction doesn't take snapshot S into account. When it reached "foo", transaction T is committed. Compaction may think the value "foo=bar" is not visible by any snapshot (which is wrong), and compact the value out.
The fix is to explicitly take a snapshot before compaction grabbing the list of snapshots. Compaction will then has to keep keys visible to this snapshot.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4883
Differential Revision: D13668775
Pulled By: maysamyabandeh
fbshipit-source-id: 1cab9615f94b7d3e8522cc3d44c3a14c7d4720e4
Summary:
The AdvanceMaxEvictedSeq algorithm assumes that new snapshots always have sequence number larger than the last max_evicted_seq_. To enforce this assumption we make two changes:
i) max is not advanced beyond the last published seq, with the exception that the evicted commit entry itself is not published yet, which is quite rare.
ii) When obtaining the snapshot if the max_evicted_seq_ is not published yet, commit a dummy entry so that it waits for it to be published and also increased the latest published seq by one above the max.
To test these non-realistic corner cases we create a commit cache with size 1 so that every single commit results into eviction.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4886
Differential Revision: D13685270
Pulled By: maysamyabandeh
fbshipit-source-id: 5461bc09c2a9b75798bfcb9853a256c81cdac0b0
Summary:
Right now the error mesage when options.wal_dir doesn't exist is not helpful to users. Be more specific
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4874
Differential Revision: D13642425
Pulled By: siying
fbshipit-source-id: 9a3172ed0f799af233b0f3b2e5e35bc7ce04c7b5
Summary:
If we do not do this, then reading MutableCFOptions may have a race condition
with SetOptions which modifies MutableCFOptions.
Also reserve space in advance for vectors to avoid reallocation changing the
address of its elements.
Test plan
```
$make clean && make -j32 all check
$make clean && COMPILE_WITH_TSAN=1 make -j32 all check
$make clean && COMPILE_WITH_ASAN=1 make -j32 all check
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4876
Differential Revision: D13644500
Pulled By: riversand963
fbshipit-source-id: 4b8112c5c819d5a2922bb61ad1521b3d2fb2fd47
Summary:
The vector returned by SnapshotList::GetAll could have duplicate entries if two separate snapshots have the same sequence number. However, when this vector is used in compaction the duplicate entires are of no use and could be safely ignored. Moreover not having duplicate entires simplifies reasoning in the compaction_iterator.cc code. For example when searching for the previous_snap we currently use the snapshot before the current one but the way the code uses that it expects it to be also less than the current snapshot, which would be simpler to read if there is no duplicate entry in the snapshot list.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4860
Differential Revision: D13615502
Pulled By: maysamyabandeh
fbshipit-source-id: d45bf01213ead5f39db811f951802da6fcc3332b
Summary:
https://github.com/facebook/rocksdb/pull/3340 introduces preloading when max_open_files != -1.
It doesn't preload index and filter in non-initial file loading case. This is a little bit too
complicated to understand. We observed in one MyRocks use case where the filter is expected to be
preloaded but is not. To simplify the use case, we simply always prefetch the index and filter.
They anyway is expected to be loaded in the file verification phase anyway.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4852
Differential Revision: D13595402
Pulled By: siying
fbshipit-source-id: d4d8624eb3e849e20aeb990df2100502d85aff31
Michael Liu