Faster new DynamicBloom implementation (for memtable) (#5762)

Summary:
Since DynamicBloom is now only used in-memory, we're free to
change it without schema compatibility issues. The new implementation
is drawn from (with manifest permission)
303542a767/bloom_simulation_tests/foo.cc (L613)

This has several speed advantages over the prior implementation:
* Uses fastrange instead of %
* Minimum logic to determine first (and all) probed memory addresses
* (Major) Two probes per 64-bit memory fetch/write.
* Very fast and effective (murmur-like) hash expansion/re-mixing. (At
least on recent CPUs, integer multiplication is very cheap.)

While a Bloom filter with 512-bit cache locality has about a 1.15x FP
rate penalty (e.g. 0.84% to 0.97%), further restricting to two probes
per 64 bits incurs an additional 1.12x FP rate penalty (e.g. 0.97% to
1.09%). Nevertheless, the unit tests show no "mediocre" FP rate samples,
unlike the old implementation with more erratic FP rates.

Especially for the memtable, we expect speed to outweigh somewhat higher
FP rates. For example, a negative table query would have to be 1000x
slower than a BF query to justify doubling BF query time to shave 10% off
FP rate (working assumption around 1% FP rate). While that seems likely
for SSTs, my data suggests a speed factor of roughly 50x for the memtable
(vs. BF; ~1.5% lower write throughput when enabling memtable Bloom
filter, after this change).  Thus, it's probably not worth even 5% more
time in the Bloom filter to shave off 1/10th of the Bloom FP rate, or 0.1%
in absolute terms, and it's probably at least 20% slower to recoup that
much FP rate from this new implementation. Because of this, we do not see
a need for a 'locality' option that affects the MemTable Bloom filter
and have decoupled the MemTable Bloom filter from Options::bloom_locality.

Note that just 3% more memory to the Bloom filter (10.3 bits per key vs.
just 10) is able to make up for the ~12% FP rate drop in the new
implementation:

[] # Nearly "ideal" FP-wise but reasonably fast cache-local implementation
[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_WORM64_FROM32_any.out 10000000 6 10 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_WORM64_FROM32_any.out time: 3.29372 sampled_fp_rate: 0.00985956 ...

[] # Close match to this new implementation
[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_MUL64_BLOCK_FROM32_any.out 10000000 6 10.3 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_MUL64_BLOCK_FROM32_any.out time: 2.10072 sampled_fp_rate: 0.00985655 ...

[] # Old locality=1 implementation
[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_ROCKSDB_DYNAMIC_any.out 10000000 6 10 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_ROCKSDB_DYNAMIC_any.out time: 3.95472 sampled_fp_rate: 0.00988943 ...

Also note the dramatic speed improvement vs. alternatives.

--

Performance unit test: DynamicBloomTest.concurrent_with_perf is updated
to report more precise timing data. (Measure running time of each
thread, not just longest running thread, etc.) Results averaged over
various sizes enabled with --enable_perf and 20 runs each; old dynamic
bloom refers to locality=1, the faster of the old:

old dynamic bloom, avg add latency = 65.6468
new dynamic bloom, avg add latency = 44.3809
old dynamic bloom, avg query latency = 50.6485
new dynamic bloom, avg query latency = 43.2186
old avg parallel add latency = 41.678
new avg parallel add latency = 24.5238
old avg parallel hit latency = 14.6322
new avg parallel hit latency = 12.3939
old avg parallel miss latency = 16.7289
new avg parallel miss latency = 12.2134

Tested on a dedicated 64-bit production machine at Facebook. Significant
improvement all around.

Despite now using std::atomic<uint64_t>, quick before-and-after test on
a 32-bit machine (Intel Atom N270, released 2008) shows no regression in
performance, in some cases modest improvement.

--

Performance integration test (synthetic): with DEBUG_LEVEL=0, used
TEST_TMPDIR=/dev/shm ./db_bench --benchmarks=fillrandom,readmissing,readrandom,stats --num=2000000
and optionally with -memtable_whole_key_filtering -memtable_bloom_size_ratio=0.01
300 runs each configuration.

Write throughput change by enabling memtable bloom:
Old locality=0: -3.06%
Old locality=1: -2.37%
New:            -1.50%
conclusion -> seems to substantially close the gap

Readmissing throughput change by enabling memtable bloom:
Old locality=0: +34.47%
Old locality=1: +34.80%
New:            +33.25%
conclusion -> maybe a small new penalty from FP rate

Readrandom throughput change by enabling memtable bloom:
Old locality=0: +31.54%
Old locality=1: +31.13%
New:            +30.60%
conclusion -> maybe also from FP rate (after memtable flush)

--

Another conclusion we can draw from this new implementation is that the
existing 32-bit hash function is not inherently crippling the Bloom
filter speed or accuracy, below about 5 million keys. For speed, the
implementation is essentially the same whether starting with 32-bits or
64-bits of hash; it just determines whether the first multiplication
after fastrange is a pseudorandom expansion or needed re-mix. Note that
this multiplication can occur while memory is fetching.

For accuracy, in a standard configuration, you need about 5 million
keys before you have about a 1.1x FP penalty due to using a
32-bit hash vs. 64-bit:

[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_MUL64_BLOCK_FROM32_any.out $((5 * 1000 * 1000 * 10)) 6 10 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_MUL64_BLOCK_FROM32_any.out time: 2.52069 sampled_fp_rate: 0.0118267 ...
[~/wormhashing/bloom_simulation_tests] ./foo_gcc_IMPL_CACHE_MUL64_BLOCK_any.out $((5 * 1000 * 1000 * 10)) 6 10 $RANDOM 100000000
./foo_gcc_IMPL_CACHE_MUL64_BLOCK_any.out time: 2.43871 sampled_fp_rate: 0.0109059
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5762

Differential Revision: D17214194

Pulled By: pdillinger

fbshipit-source-id: ad9da031772e985fd6b62a0e1db8e81892520595
This commit is contained in:
Peter Dillinger 2019-09-05 14:57:39 -07:00 committed by Facebook Github Bot
parent 19e8c9b64f
commit b55b2f45d0
7 changed files with 235 additions and 244 deletions

View File

@ -11,6 +11,9 @@
### Public API Change
* Added max_write_buffer_size_to_maintain option to better control memory usage of immutable memtables.
* Added a lightweight API GetCurrentWalFile() to get last live WAL filename and size. Meant to be used as a helper for backup/restore tooling in a larger ecosystem such as MySQL with a MyRocks storage engine.
* The MemTable Bloom filter, when enabled, now always uses cache locality. Options::bloom_locality now only affects the PlainTable SST format.
### Performance Improvements
* Improve the speed of the MemTable Bloom filter, reducing the write overhead of enabling it by 1/3 to 1/2, with similar benefit to read performance.
## 6.4.0 (7/30/2019)
### Default Option Change

View File

@ -116,7 +116,7 @@ MemTable::MemTable(const InternalKeyComparator& cmp,
moptions_.memtable_prefix_bloom_bits > 0) {
bloom_filter_.reset(
new DynamicBloom(&arena_, moptions_.memtable_prefix_bloom_bits,
ioptions.bloom_locality, 6 /* hard coded 6 probes */,
6 /* hard coded 6 probes */,
moptions_.memtable_huge_page_size, ioptions.info_log));
}
}

View File

@ -19,6 +19,8 @@ class Slice;
class Allocator;
class Logger;
// A legacy Bloom filter implementation used by Plain Table db format, for
// schema backward compatibility. Not for use in new filter applications.
class PlainTableBloomV1 {
public:
// allocator: pass allocator to bloom filter, hence trace the usage of memory

View File

@ -16,44 +16,49 @@ namespace rocksdb {
namespace {
uint32_t GetTotalBitsForLocality(uint32_t total_bits) {
uint32_t num_blocks =
(total_bits + CACHE_LINE_SIZE * 8 - 1) / (CACHE_LINE_SIZE * 8);
// Make num_blocks an odd number to make sure more bits are involved
// when determining which block.
if (num_blocks % 2 == 0) {
num_blocks++;
uint32_t roundUpToPow2(uint32_t x) {
uint32_t rv = 1;
while (rv < x) {
rv <<= 1;
}
return num_blocks * (CACHE_LINE_SIZE * 8);
return rv;
}
}
DynamicBloom::DynamicBloom(Allocator* allocator, uint32_t total_bits,
uint32_t locality, uint32_t num_probes,
uint32_t num_probes,
size_t huge_page_tlb_size, Logger* logger)
: kNumProbes(num_probes) {
kTotalBits = (locality > 0) ? GetTotalBitsForLocality(total_bits)
: (total_bits + 7) / 8 * 8;
kNumBlocks = (locality > 0) ? (kTotalBits / (CACHE_LINE_SIZE * 8)) : 0;
// Round down, except round up with 1
: kNumDoubleProbes((num_probes + (num_probes == 1)) / 2) {
assert(num_probes % 2 == 0); // limitation of current implementation
assert(num_probes <= 10); // limitation of current implementation
assert(kNumDoubleProbes > 0);
assert(kNumBlocks > 0 || kTotalBits > 0);
assert(kNumProbes > 0);
// Determine how much to round off + align by so that x ^ i (that's xor) is
// a valid u64 index if x is a valid u64 index and 0 <= i < kNumDoubleProbes.
uint32_t block_bytes = /*bytes/u64*/ 8 *
/*u64s*/ std::max(1U, roundUpToPow2(kNumDoubleProbes));
kLen = (total_bits + (/*bits/byte*/ 8 * block_bytes - 1)) /
/*bits/u64*/ 64;
assert(kLen > 0);
uint32_t sz = kTotalBits / 8;
if (kNumBlocks > 0) {
sz += CACHE_LINE_SIZE - 1;
}
uint32_t sz = kLen * /*bytes/u64*/ 8;
// Padding to correct for allocation not originally aligned on block_bytes
// boundary
sz += block_bytes - 1;
assert(allocator);
char* raw = allocator->AllocateAligned(sz, huge_page_tlb_size, logger);
memset(raw, 0, sz);
auto cache_line_offset = reinterpret_cast<uintptr_t>(raw) % CACHE_LINE_SIZE;
if (kNumBlocks > 0 && cache_line_offset > 0) {
raw += CACHE_LINE_SIZE - cache_line_offset;
auto block_offset = reinterpret_cast<uintptr_t>(raw) % block_bytes;
if (block_offset > 0) {
// Align on block_bytes boundary
raw += block_bytes - block_offset;
}
data_ = reinterpret_cast<std::atomic<uint8_t>*>(raw);
static_assert(sizeof(std::atomic<uint64_t>) == sizeof(uint64_t),
"Expecting zero-space-overhead atomic");
data_ = reinterpret_cast<std::atomic<uint64_t>*>(raw);
}
} // rocksdb

View File

@ -24,12 +24,20 @@ class Logger;
// A Bloom filter intended only to be used in memory, never serialized in a way
// that could lead to schema incompatibility. Supports opt-in lock-free
// concurrent access.
//
// This implementation is also intended for applications generally preferring
// speed vs. maximum accuracy: roughly 0.9x BF op latency for 1.1x FP rate.
// For 1% FP rate, that means that the latency of a look-up triggered by an FP
// should be less than roughly 100x the cost of a Bloom filter op.
//
// For simplicity and performance, the current implementation requires
// num_probes to be a multiple of two and <= 10.
//
class DynamicBloom {
public:
// allocator: pass allocator to bloom filter, hence trace the usage of memory
// total_bits: fixed total bits for the bloom
// num_probes: number of hash probes for a single key
// locality: If positive, optimize for cache line locality, 0 otherwise.
// hash_func: customized hash function
// huge_page_tlb_size: if >0, try to allocate bloom bytes from huge page TLB
// within this page size. Need to reserve huge pages for
@ -37,7 +45,7 @@ class DynamicBloom {
// sysctl -w vm.nr_hugepages=20
// See linux doc Documentation/vm/hugetlbpage.txt
explicit DynamicBloom(Allocator* allocator,
uint32_t total_bits, uint32_t locality = 0,
uint32_t total_bits,
uint32_t num_probes = 6,
size_t huge_page_tlb_size = 0,
Logger* logger = nullptr);
@ -64,14 +72,15 @@ class DynamicBloom {
void Prefetch(uint32_t h);
uint32_t GetNumBlocks() const { return kNumBlocks; }
private:
uint32_t kTotalBits;
uint32_t kNumBlocks;
const uint32_t kNumProbes;
// Length of the structure, in 64-bit words. For this structure, "word"
// will always refer to 64-bit words.
uint32_t kLen;
// We make the k probes in pairs, two for each 64-bit read/write. Thus,
// this stores k/2, the number of words to double-probe.
const uint32_t kNumDoubleProbes;
std::atomic<uint8_t>* data_;
std::atomic<uint64_t>* data_;
// or_func(ptr, mask) should effect *ptr |= mask with the appropriate
// concurrency safety, working with bytes.
@ -86,14 +95,14 @@ inline void DynamicBloom::AddConcurrently(const Slice& key) {
}
inline void DynamicBloom::AddHash(uint32_t hash) {
AddHash(hash, [](std::atomic<uint8_t>* ptr, uint8_t mask) {
AddHash(hash, [](std::atomic<uint64_t>* ptr, uint64_t mask) {
ptr->store(ptr->load(std::memory_order_relaxed) | mask,
std::memory_order_relaxed);
});
}
inline void DynamicBloom::AddHashConcurrently(uint32_t hash) {
AddHash(hash, [](std::atomic<uint8_t>* ptr, uint8_t mask) {
AddHash(hash, [](std::atomic<uint64_t>* ptr, uint64_t mask) {
// Happens-before between AddHash and MaybeContains is handled by
// access to versions_->LastSequence(), so all we have to do here is
// avoid races (so we don't give the compiler a license to mess up
@ -114,67 +123,69 @@ inline bool DynamicBloom::MayContain(const Slice& key) const {
// local variable is initialized but not referenced
#pragma warning(disable : 4189)
#endif
inline void DynamicBloom::Prefetch(uint32_t h) {
if (kNumBlocks != 0) {
uint32_t b = ((h >> 11 | (h << 21)) % kNumBlocks) * (CACHE_LINE_SIZE * 8);
PREFETCH(&(data_[b / 8]), 0, 3);
}
inline void DynamicBloom::Prefetch(uint32_t h32) {
size_t a = fastrange32(kLen, h32);
PREFETCH(data_ + a, 0, 3);
}
#if defined(_MSC_VER)
#pragma warning(pop)
#endif
inline bool DynamicBloom::MayContainHash(uint32_t h) const {
const uint32_t delta = (h >> 17) | (h << 15); // Rotate right 17 bits
if (kNumBlocks != 0) {
uint32_t b = ((h >> 11 | (h << 21)) % kNumBlocks) * (CACHE_LINE_SIZE * 8);
for (uint32_t i = 0; i < kNumProbes; ++i) {
// Since CACHE_LINE_SIZE is defined as 2^n, this line will be optimized
// to a simple and operation by compiler.
const uint32_t bitpos = b + (h % (CACHE_LINE_SIZE * 8));
uint8_t byteval = data_[bitpos / 8].load(std::memory_order_relaxed);
if ((byteval & (1 << (bitpos % 8))) == 0) {
return false;
}
// Rotate h so that we don't reuse the same bytes.
h = h / (CACHE_LINE_SIZE * 8) +
(h % (CACHE_LINE_SIZE * 8)) * (0x20000000U / CACHE_LINE_SIZE);
h += delta;
}
} else {
for (uint32_t i = 0; i < kNumProbes; ++i) {
const uint32_t bitpos = h % kTotalBits;
uint8_t byteval = data_[bitpos / 8].load(std::memory_order_relaxed);
if ((byteval & (1 << (bitpos % 8))) == 0) {
return false;
}
h += delta;
// Speed hacks in this implementation:
// * Uses fastrange instead of %
// * Minimum logic to determine first (and all) probed memory addresses.
// (Uses constant bit-xor offsets from the starting probe address.)
// * (Major) Two probes per 64-bit memory fetch/write.
// Code simplification / optimization: only allow even number of probes.
// * Very fast and effective (murmur-like) hash expansion/re-mixing. (At
// least on recent CPUs, integer multiplication is very cheap. Each 64-bit
// remix provides five pairs of bit addresses within a uint64_t.)
// Code simplification / optimization: only allow up to 10 probes, from a
// single 64-bit remix.
//
// The FP rate penalty for this implementation, vs. standard Bloom filter, is
// roughly 1.12x on top of the 1.15x penalty for a 512-bit cache-local Bloom.
// This implementation does not explicitly use the cache line size, but is
// effectively cache-local (up to 16 probes) because of the bit-xor offsetting.
//
// NB: could easily be upgraded to support a 64-bit hash and
// total_bits > 2^32 (512MB). (The latter is a bad idea without the former,
// because of false positives.)
inline bool DynamicBloom::MayContainHash(uint32_t h32) const {
size_t a = fastrange32(kLen, h32);
PREFETCH(data_ + a, 0, 3);
// Expand/remix with 64-bit golden ratio
uint64_t h = 0x9e3779b97f4a7c13ULL * h32;
for (unsigned i = 0;; ++i) {
// Two bit probes per uint64_t probe
uint64_t mask = ((uint64_t)1 << (h & 63))
| ((uint64_t)1 << ((h >> 6) & 63));
uint64_t val = data_[a ^ i].load(std::memory_order_relaxed);
if (i + 1 >= kNumDoubleProbes) {
return (val & mask) == mask;
} else if ((val & mask) != mask) {
return false;
}
h = (h >> 12) | (h << 52);
}
return true;
}
template <typename OrFunc>
inline void DynamicBloom::AddHash(uint32_t h, const OrFunc& or_func) {
const uint32_t delta = (h >> 17) | (h << 15); // Rotate right 17 bits
if (kNumBlocks != 0) {
uint32_t b = ((h >> 11 | (h << 21)) % kNumBlocks) * (CACHE_LINE_SIZE * 8);
for (uint32_t i = 0; i < kNumProbes; ++i) {
// Since CACHE_LINE_SIZE is defined as 2^n, this line will be optimized
// to a simple and operation by compiler.
const uint32_t bitpos = b + (h % (CACHE_LINE_SIZE * 8));
or_func(&data_[bitpos / 8], (1 << (bitpos % 8)));
// Rotate h so that we don't reuse the same bytes.
h = h / (CACHE_LINE_SIZE * 8) +
(h % (CACHE_LINE_SIZE * 8)) * (0x20000000U / CACHE_LINE_SIZE);
h += delta;
}
} else {
for (uint32_t i = 0; i < kNumProbes; ++i) {
const uint32_t bitpos = h % kTotalBits;
or_func(&data_[bitpos / 8], (1 << (bitpos % 8)));
h += delta;
inline void DynamicBloom::AddHash(uint32_t h32, const OrFunc& or_func) {
size_t a = fastrange32(kLen, h32);
PREFETCH(data_ + a, 0, 3);
// Expand/remix with 64-bit golden ratio
uint64_t h = 0x9e3779b97f4a7c13ULL * h32;
for (unsigned i = 0;; ++i) {
// Two bit probes per uint64_t probe
uint64_t mask = ((uint64_t)1 << (h & 63))
| ((uint64_t)1 << ((h >> 6) & 63));
or_func(&data_[a ^ i], mask);
if (i + 1 >= kNumDoubleProbes) {
return;
}
h = (h >> 12) | (h << 52);
}
}

View File

@ -45,18 +45,18 @@ class DynamicBloomTest : public testing::Test {};
TEST_F(DynamicBloomTest, EmptyFilter) {
Arena arena;
DynamicBloom bloom1(&arena, 100, 0, 2);
DynamicBloom bloom1(&arena, 100, 2);
ASSERT_TRUE(!bloom1.MayContain("hello"));
ASSERT_TRUE(!bloom1.MayContain("world"));
DynamicBloom bloom2(&arena, CACHE_LINE_SIZE * 8 * 2 - 1, 1, 2);
DynamicBloom bloom2(&arena, CACHE_LINE_SIZE * 8 * 2 - 1, 2);
ASSERT_TRUE(!bloom2.MayContain("hello"));
ASSERT_TRUE(!bloom2.MayContain("world"));
}
TEST_F(DynamicBloomTest, Small) {
Arena arena;
DynamicBloom bloom1(&arena, 100, 0, 2);
DynamicBloom bloom1(&arena, 100, 2);
bloom1.Add("hello");
bloom1.Add("world");
ASSERT_TRUE(bloom1.MayContain("hello"));
@ -64,7 +64,7 @@ TEST_F(DynamicBloomTest, Small) {
ASSERT_TRUE(!bloom1.MayContain("x"));
ASSERT_TRUE(!bloom1.MayContain("foo"));
DynamicBloom bloom2(&arena, CACHE_LINE_SIZE * 8 * 2 - 1, 1, 2);
DynamicBloom bloom2(&arena, CACHE_LINE_SIZE * 8 * 2 - 1, 2);
bloom2.Add("hello");
bloom2.Add("world");
ASSERT_TRUE(bloom2.MayContain("hello"));
@ -75,7 +75,7 @@ TEST_F(DynamicBloomTest, Small) {
TEST_F(DynamicBloomTest, SmallConcurrentAdd) {
Arena arena;
DynamicBloom bloom1(&arena, 100, 0, 2);
DynamicBloom bloom1(&arena, 100, 2);
bloom1.AddConcurrently("hello");
bloom1.AddConcurrently("world");
ASSERT_TRUE(bloom1.MayContain("hello"));
@ -83,7 +83,7 @@ TEST_F(DynamicBloomTest, SmallConcurrentAdd) {
ASSERT_TRUE(!bloom1.MayContain("x"));
ASSERT_TRUE(!bloom1.MayContain("foo"));
DynamicBloom bloom2(&arena, CACHE_LINE_SIZE * 8 * 2 - 1, 1, 2);
DynamicBloom bloom2(&arena, CACHE_LINE_SIZE * 8 * 2 - 1, 2);
bloom2.AddConcurrently("hello");
bloom2.AddConcurrently("world");
ASSERT_TRUE(bloom2.MayContain("hello"));
@ -116,53 +116,44 @@ TEST_F(DynamicBloomTest, VaryingLengths) {
fprintf(stderr, "bits_per_key: %d num_probes: %d\n", FLAGS_bits_per_key,
num_probes);
for (uint32_t enable_locality = 0; enable_locality < 2; ++enable_locality) {
for (uint32_t num = 1; num <= 10000; num = NextNum(num)) {
uint32_t bloom_bits = 0;
Arena arena;
if (enable_locality == 0) {
bloom_bits = std::max(num * FLAGS_bits_per_key, 64U);
} else {
bloom_bits = std::max(num * FLAGS_bits_per_key,
enable_locality * CACHE_LINE_SIZE * 8);
}
DynamicBloom bloom(&arena, bloom_bits, enable_locality, num_probes);
for (uint64_t i = 0; i < num; i++) {
bloom.Add(Key(i, buffer));
ASSERT_TRUE(bloom.MayContain(Key(i, buffer)));
}
// All added keys must match
for (uint64_t i = 0; i < num; i++) {
ASSERT_TRUE(bloom.MayContain(Key(i, buffer))) << "Num " << num
<< "; key " << i;
}
// Check false positive rate
int result = 0;
for (uint64_t i = 0; i < 10000; i++) {
if (bloom.MayContain(Key(i + 1000000000, buffer))) {
result++;
}
}
double rate = result / 10000.0;
fprintf(stderr,
"False positives: %5.2f%% @ num = %6u, bloom_bits = %6u, "
"enable locality?%u\n",
rate * 100.0, num, bloom_bits, enable_locality);
if (rate > 0.0125)
mediocre_filters++; // Allowed, but not too often
else
good_filters++;
for (uint32_t num = 1; num <= 10000; num = NextNum(num)) {
uint32_t bloom_bits = 0;
Arena arena;
bloom_bits = num * FLAGS_bits_per_key;
DynamicBloom bloom(&arena, bloom_bits, num_probes);
for (uint64_t i = 0; i < num; i++) {
bloom.Add(Key(i, buffer));
ASSERT_TRUE(bloom.MayContain(Key(i, buffer)));
}
fprintf(stderr, "Filters: %d good, %d mediocre\n", good_filters,
mediocre_filters);
ASSERT_LE(mediocre_filters, good_filters / 5);
// All added keys must match
for (uint64_t i = 0; i < num; i++) {
ASSERT_TRUE(bloom.MayContain(Key(i, buffer))) << "Num " << num
<< "; key " << i;
}
// Check false positive rate
int result = 0;
for (uint64_t i = 0; i < 10000; i++) {
if (bloom.MayContain(Key(i + 1000000000, buffer))) {
result++;
}
}
double rate = result / 10000.0;
fprintf(stderr,
"False positives: %5.2f%% @ num = %6u, bloom_bits = %6u\n",
rate * 100.0, num, bloom_bits);
if (rate > 0.0125)
mediocre_filters++; // Allowed, but not too often
else
good_filters++;
}
fprintf(stderr, "Filters: %d good, %d mediocre\n", good_filters,
mediocre_filters);
ASSERT_LE(mediocre_filters, good_filters / 5);
}
TEST_F(DynamicBloomTest, perf) {
@ -178,7 +169,7 @@ TEST_F(DynamicBloomTest, perf) {
const uint32_t num_keys = m * 8 * 1024 * 1024;
fprintf(stderr, "testing %" PRIu32 "M keys\n", m * 8);
DynamicBloom std_bloom(&arena, num_keys * 10, 0, num_probes);
DynamicBloom std_bloom(&arena, num_keys * 10, num_probes);
timer.Start();
for (uint64_t i = 1; i <= num_keys; ++i) {
@ -186,8 +177,8 @@ TEST_F(DynamicBloomTest, perf) {
}
uint64_t elapsed = timer.ElapsedNanos();
fprintf(stderr, "standard bloom, avg add latency %" PRIu64 "\n",
elapsed / num_keys);
fprintf(stderr, "dynamic bloom, avg add latency %3g\n",
static_cast<double>(elapsed) / num_keys);
uint32_t count = 0;
timer.Start();
@ -199,128 +190,99 @@ TEST_F(DynamicBloomTest, perf) {
ASSERT_EQ(count, num_keys);
elapsed = timer.ElapsedNanos();
assert(count > 0);
fprintf(stderr, "standard bloom, avg query latency %" PRIu64 "\n",
elapsed / count);
// Locality enabled version
DynamicBloom blocked_bloom(&arena, num_keys * 10, 1, num_probes);
timer.Start();
for (uint64_t i = 1; i <= num_keys; ++i) {
blocked_bloom.Add(Slice(reinterpret_cast<const char*>(&i), 8));
}
elapsed = timer.ElapsedNanos();
fprintf(stderr,
"blocked bloom(enable locality), avg add latency %" PRIu64 "\n",
elapsed / num_keys);
count = 0;
timer.Start();
for (uint64_t i = 1; i <= num_keys; ++i) {
if (blocked_bloom.MayContain(
Slice(reinterpret_cast<const char*>(&i), 8))) {
++count;
}
}
elapsed = timer.ElapsedNanos();
assert(count > 0);
fprintf(stderr,
"blocked bloom(enable locality), avg query latency %" PRIu64 "\n",
elapsed / count);
ASSERT_TRUE(count == num_keys);
fprintf(stderr, "dynamic bloom, avg query latency %3g\n",
static_cast<double>(elapsed) / count);
}
}
TEST_F(DynamicBloomTest, concurrent_with_perf) {
StopWatchNano timer(Env::Default());
uint32_t num_probes = static_cast<uint32_t>(FLAGS_num_probes);
uint32_t m_limit = FLAGS_enable_perf ? 8 : 1;
uint32_t locality_limit = FLAGS_enable_perf ? 1 : 0;
uint32_t num_threads = 4;
std::vector<port::Thread> threads;
for (uint32_t m = 1; m <= m_limit; ++m) {
for (uint32_t locality = 0; locality <= locality_limit; ++locality) {
Arena arena;
const uint32_t num_keys = m * 8 * 1024 * 1024;
fprintf(stderr, "testing %" PRIu32 "M keys with %" PRIu32 " locality\n",
m * 8, locality);
Arena arena;
const uint32_t num_keys = m * 8 * 1024 * 1024;
fprintf(stderr, "testing %" PRIu32 "M keys\n", m * 8);
DynamicBloom std_bloom(&arena, num_keys * 10, locality, num_probes);
DynamicBloom std_bloom(&arena, num_keys * 10, num_probes);
std::atomic<uint64_t> elapsed(0);
std::function<void(size_t)> adder([&](size_t t) {
StopWatchNano timer(Env::Default());
timer.Start();
std::function<void(size_t)> adder([&](size_t t) {
for (uint64_t i = 1 + t; i <= num_keys; i += num_threads) {
std_bloom.AddConcurrently(
Slice(reinterpret_cast<const char*>(&i), 8));
}
});
for (size_t t = 0; t < num_threads; ++t) {
threads.emplace_back(adder, t);
for (uint64_t i = 1 + t; i <= num_keys; i += num_threads) {
std_bloom.AddConcurrently(
Slice(reinterpret_cast<const char*>(&i), 8));
}
while (threads.size() > 0) {
threads.back().join();
threads.pop_back();
}
uint64_t elapsed = timer.ElapsedNanos();
fprintf(stderr, "standard bloom, avg parallel add latency %" PRIu64
" nanos/key\n",
elapsed / num_keys);
timer.Start();
std::function<void(size_t)> hitter([&](size_t t) {
for (uint64_t i = 1 + t; i <= num_keys; i += num_threads) {
bool f =
std_bloom.MayContain(Slice(reinterpret_cast<const char*>(&i), 8));
ASSERT_TRUE(f);
}
});
for (size_t t = 0; t < num_threads; ++t) {
threads.emplace_back(hitter, t);
}
while (threads.size() > 0) {
threads.back().join();
threads.pop_back();
}
elapsed = timer.ElapsedNanos();
fprintf(stderr, "standard bloom, avg parallel hit latency %" PRIu64
" nanos/key\n",
elapsed / num_keys);
timer.Start();
std::atomic<uint32_t> false_positives(0);
std::function<void(size_t)> misser([&](size_t t) {
for (uint64_t i = num_keys + 1 + t; i <= 2 * num_keys;
i += num_threads) {
bool f =
std_bloom.MayContain(Slice(reinterpret_cast<const char*>(&i), 8));
if (f) {
++false_positives;
}
}
});
for (size_t t = 0; t < num_threads; ++t) {
threads.emplace_back(misser, t);
}
while (threads.size() > 0) {
threads.back().join();
threads.pop_back();
}
elapsed = timer.ElapsedNanos();
fprintf(stderr, "standard bloom, avg parallel miss latency %" PRIu64
" nanos/key, %f%% false positive rate\n",
elapsed / num_keys, false_positives.load() * 100.0 / num_keys);
elapsed += timer.ElapsedNanos();
});
for (size_t t = 0; t < num_threads; ++t) {
threads.emplace_back(adder, t);
}
while (threads.size() > 0) {
threads.back().join();
threads.pop_back();
}
fprintf(stderr, "dynamic bloom, avg parallel add latency %3g"
" nanos/key\n",
static_cast<double>(elapsed) / num_threads / num_keys);
elapsed = 0;
std::function<void(size_t)> hitter([&](size_t t) {
StopWatchNano timer(Env::Default());
timer.Start();
for (uint64_t i = 1 + t; i <= num_keys; i += num_threads) {
bool f =
std_bloom.MayContain(Slice(reinterpret_cast<const char*>(&i), 8));
ASSERT_TRUE(f);
}
elapsed += timer.ElapsedNanos();
});
for (size_t t = 0; t < num_threads; ++t) {
threads.emplace_back(hitter, t);
}
while (threads.size() > 0) {
threads.back().join();
threads.pop_back();
}
fprintf(stderr, "dynamic bloom, avg parallel hit latency %3g"
" nanos/key\n",
static_cast<double>(elapsed) / num_threads / num_keys);
elapsed = 0;
std::atomic<uint32_t> false_positives(0);
std::function<void(size_t)> misser([&](size_t t) {
StopWatchNano timer(Env::Default());
timer.Start();
for (uint64_t i = num_keys + 1 + t; i <= 2 * num_keys;
i += num_threads) {
bool f =
std_bloom.MayContain(Slice(reinterpret_cast<const char*>(&i), 8));
if (f) {
++false_positives;
}
}
elapsed += timer.ElapsedNanos();
});
for (size_t t = 0; t < num_threads; ++t) {
threads.emplace_back(misser, t);
}
while (threads.size() > 0) {
threads.back().join();
threads.pop_back();
}
fprintf(stderr, "dynamic bloom, avg parallel miss latency %3g"
" nanos/key, %f%% false positive rate\n",
static_cast<double>(elapsed) / num_threads / num_keys,
false_positives.load() * 100.0 / num_keys);
}
}

View File

@ -49,4 +49,12 @@ struct SliceHasher {
uint32_t operator()(const Slice& s) const { return GetSliceHash(s); }
};
// An alternative to % for mapping a hash value to an arbitrary range. See
// https://github.com/lemire/fastrange and
// https://github.com/pdillinger/wormhashing/blob/2c4035a4462194bf15f3e9fc180c27c513335225/bloom_simulation_tests/foo.cc#L57
inline uint32_t fastrange32(uint32_t a, uint32_t h) {
uint64_t product = static_cast<uint64_t>(a) * h;
return static_cast<uint32_t>(product >> 32);
}
} // namespace rocksdb