Summary:
The motivations for this change include
* Free up space in ClockHandle so that we can add data for secondary cache handling while still keeping within single cache line (64 byte) size.
* This change frees up space by eliminating the need for the `hash` field by making the fixed-size key itself a hash, using a 128-bit bijective (lossless) hash.
* Generally more customizability of ShardedCache (such as hashing) without worrying about virtual call overheads
* ShardedCache now uses static polymorphism (template) instead of dynamic polymorphism (virtual overrides) for the CacheShard. No obvious performance benefit is seen from the change (as mostly expected; most calls to virtual functions in CacheShard could already be optimized to static calls), but offers more flexibility without incurring the runtime cost of adhering to a common interface (without type parameters or static callbacks).
* You'll also notice less `reinterpret_cast`ing and other boilerplate in the Cache implementations, as this can go in ShardedCache.
More detail:
* Don't have LRUCacheShard maintain `std::shared_ptr<SecondaryCache>` copies (extra refcount) when LRUCache can be in charge of keeping a `shared_ptr`.
* Renamed `capacity_mutex_` to `config_mutex_` to better represent the scope of what it guards.
* Some preparation for 64-bit hash and indexing in LRUCache, but didn't include the full change because of slight performance regression.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10801
Test Plan:
Unit test updates were non-trivial because of major changes to the ClockCacheShard interface in handling of key vs. hash.
Performance:
Create with `TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=30000000 -disable_wal=1 -bloom_bits=16`
Test with
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=readrandom[-X1000] -readonly -num=30000000 -bloom_bits=16 -cache_index_and_filter_blocks=1 -cache_size=610000000 -duration 20 -threads=16
```
Before: `readrandom [AVG 150 runs] : 321147 (± 253) ops/sec`
After: `readrandom [AVG 150 runs] : 321530 (± 326) ops/sec`
So possibly ~0.1% improvement.
And with `-cache_type=hyper_clock_cache`:
Before: `readrandom [AVG 30 runs] : 614126 (± 7978) ops/sec`
After: `readrandom [AVG 30 runs] : 645349 (± 8087) ops/sec`
So roughly 5% improvement!
Reviewed By: anand1976
Differential Revision: D40252236
Pulled By: pdillinger
fbshipit-source-id: ff8fc70ef569585edc95bcbaaa0386f61355ae5b
Summary:
This is intended as a step toward possibly separating secondary cache integration from the
Cache implementation as much as possible, to (hopefully) minimize code duplication in
adding secondary cache support to HyperClockCache.
* Major clarifications to API docs of secondary cache compatible parts of Cache. For example, previously the docs seemed to suggest that Wait() was not needed if IsReady()==true. And it wasn't clear what operations were actually supported on pending handles.
* Add some assertions related to these requirements, such as that we don't Release() before Wait() (which would leak a secondary cache handle).
* Fix a leaky abstraction with dummy handles, which are supposed to be internal to the Cache. Previously, these just used value=nullptr to indicate dummy handle, which meant that they could be confused with legitimate value=nullptr cases like cache reservations. Also fixed blob_source_test which was relying on this leaky abstraction.
* Drop "incomplete" terminology, which was another name for "pending".
* Split handle flags into "mutable" ones requiring mutex and "immutable" ones which do not. Because of single-threaded access to pending handles, the "Is Pending" flag can be in the "immutable" set. This allows removal of a TSAN work-around and removing a mutex acquire-release in IsReady().
* Remove some unnecessary handling of charges on handles of failed lookups. Keeping total_charge=0 means no special handling needed. (Removed one unnecessary mutex acquire/release.)
* Simplify handling of dummy handle in Lookup(). There is no need to explicitly Ref & Release w/Erase if we generally overwrite the dummy anyway. (Removed one mutex acquire/release, a call to Release().)
Intended follow-up:
* Clarify APIs in secondary_cache.h
* Doesn't SecondaryCacheResultHandle transfer ownership of the Value() on success (implementations should not release the value in destructor)?
* Does Wait() need to be called if IsReady() == true? (This would be different from Cache.)
* Do Value() and Size() have undefined behavior if IsReady() == false?
* Why have a custom API for what is essentially a std::future<std::pair<void*, size_t>>?
* Improve unit testing of standalone handle case
* Apparent null `e` bug in `free_standalone_handle` case
* Clean up secondary cache testing in lru_cache_test
* Why does TestSecondaryCacheResultHandle hold on to a Cache::Handle?
* Why does TestSecondaryCacheResultHandle::Wait() do nothing? Shouldn't it establish the post-condition IsReady() == true?
* (Assuming that is sorted out...) Shouldn't TestSecondaryCache::WaitAll simply wait on each handle in order (no casting required)? How about making that the default implementation?
* Why does TestSecondaryCacheResultHandle::Size() check Value() first? If the API is intended to be returning 0 before IsReady(), then that is weird but should at least be documented. Otherwise, if it's intended to be undefined behavior, we should assert IsReady().
* Consider replacing "standalone" and "dummy" entries with a single kind of "weak" entry that deletes its value when it reaches zero refs. Suppose you are using compressed secondary cache and have two iterators at similar places. It will probably common for one iterator to have standalone results pinned (out of cache) when the second iterator needs those same blocks and has to re-load them from secondary cache and duplicate the memory. Combining the dummy and the standalone should fix this.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10730
Test Plan:
existing tests (minor update), and crash test with sanitizers and secondary cache
Performance test for any regressions in LRUCache (primary only):
Create DB with
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=30000000 -disable_wal=1 -bloom_bits=16
```
Test before & after (run at same time) with
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=readrandom[-X100] -readonly -num=30000000 -bloom_bits=16 -cache_index_and_filter_blocks=1 -cache_size=233000000 -duration 30 -threads=16
```
Before: readrandom [AVG 100 runs] : 22234 (± 63) ops/sec; 1.6 (± 0.0) MB/sec
After: readrandom [AVG 100 runs] : 22197 (± 64) ops/sec; 1.6 (± 0.0) MB/sec
That's within 0.2%, which is not significant by the confidence intervals.
Reviewed By: anand1976
Differential Revision: D39826010
Pulled By: anand1976
fbshipit-source-id: 3202b4a91f673231c97648ae070e502ae16b0f44
Summary:
* Consolidates most metadata into a single word per slot so that more
can be accomplished with a single atomic update. In the common case,
Lookup was previously about 4 atomic updates, now just 1 atomic update.
Common case Release was previously 1 atomic read + 1 atomic update,
now just 1 atomic update.
* Eliminate spins / waits / yields, which likely threaten some "lock free"
benefits. Compare-exchange loops are only used in explicit Erase, and
strict_capacity_limit=true Insert. Eviction uses opportunistic compare-
exchange.
* Relaxes some aggressiveness and guarantees. For example,
* Duplicate Inserts will sometimes go undetected and the shadow duplicate
will age out with eviction.
* In many cases, the older Inserted value for a given cache key will be kept
(i.e. Insert does not support overwrite).
* Entries explicitly erased (rather than evicted) might not be freed
immediately in some rare cases.
* With strict_capacity_limit=false, capacity limit is not tracked/enforced as
precisely as LRUCache, but is self-correcting and should only deviate by a
very small number of extra or fewer entries.
* Use smaller "computed default" number of cache shards in many cases,
because benefits to larger usage tracking / eviction pools outweigh the small
cost of more lock-free atomic contention. The improvement in CPU and I/O
is dramatic in some limit-memory cases.
* Even without the sharding change, the eviction algorithm is likely more
effective than LRU overall because it's more stateful, even though the
"hot path" state tracking for it is essentially free with ref counting. It
is like a generalized CLOCK with aging (see code comments). I don't have
performance numbers showing a specific improvement, but in theory, for a
Poisson access pattern to each block, keeping some state allows better
estimation of time to next access (Poisson interval) than strict LRU. The
bounded randomness in CLOCK can also reduce "cliff" effect for repeated
range scans approaching and exceeding cache size.
## Hot path algorithm comparison
Rough descriptions, focusing on number and kind of atomic operations:
* Old `Lookup()` (2-5 atomic updates per probe):
```
Loop:
Increment internal ref count at slot
If possible hit:
Check flags atomic (and non-atomic fields)
If cache hit:
Three distinct updates to 'flags' atomic
Increment refs for internal-to-external
Return
Decrement internal ref count
while atomic read 'displacements' > 0
```
* New `Lookup()` (1-2 atomic updates per probe):
```
Loop:
Increment acquire counter in meta word (optimistic)
If visible entry (already read meta word):
If match (read non-atomic fields):
Return
Else:
Decrement acquire counter in meta word
Else if invisible entry (rare, already read meta word):
Decrement acquire counter in meta word
while atomic read 'displacements' > 0
```
* Old `Release()` (1 atomic update, conditional on atomic read, rarely more):
```
Read atomic ref count
If last reference and invisible (rare):
Use CAS etc. to remove
Return
Else:
Decrement ref count
```
* New `Release()` (1 unconditional atomic update, rarely more):
```
Increment release counter in meta word
If last reference and invisible (rare):
Use CAS etc. to remove
Return
```
## Performance test setup
Build DB with
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=30000000 -disable_wal=1 -bloom_bits=16
```
Test with
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=readrandom -readonly -num=30000000 -bloom_bits=16 -cache_index_and_filter_blocks=1 -cache_size=${CACHE_MB}000000 -duration 60 -threads=$THREADS -statistics
```
Numbers on a single socket Skylake Xeon system with 48 hardware threads, DEBUG_LEVEL=0 PORTABLE=0. Very similar story on a dual socket system with 80 hardware threads. Using (every 2nd) Fibonacci MB cache sizes to sample the territory between powers of two. Configurations:
base: LRUCache before this change, but with db_bench change to default cache_numshardbits=-1 (instead of fixed at 6)
folly: LRUCache before this change, with folly enabled (distributed mutex) but on an old compiler (sorry)
gt_clock: experimental ClockCache before this change
new_clock: experimental ClockCache with this change
## Performance test results
First test "hot path" read performance, with block cache large enough for whole DB:
4181MB 1thread base -> kops/s: 47.761
4181MB 1thread folly -> kops/s: 45.877
4181MB 1thread gt_clock -> kops/s: 51.092
4181MB 1thread new_clock -> kops/s: 53.944
4181MB 16thread base -> kops/s: 284.567
4181MB 16thread folly -> kops/s: 249.015
4181MB 16thread gt_clock -> kops/s: 743.762
4181MB 16thread new_clock -> kops/s: 861.821
4181MB 24thread base -> kops/s: 303.415
4181MB 24thread folly -> kops/s: 266.548
4181MB 24thread gt_clock -> kops/s: 975.706
4181MB 24thread new_clock -> kops/s: 1205.64 (~= 24 * 53.944)
4181MB 32thread base -> kops/s: 311.251
4181MB 32thread folly -> kops/s: 274.952
4181MB 32thread gt_clock -> kops/s: 1045.98
4181MB 32thread new_clock -> kops/s: 1370.38
4181MB 48thread base -> kops/s: 310.504
4181MB 48thread folly -> kops/s: 268.322
4181MB 48thread gt_clock -> kops/s: 1195.65
4181MB 48thread new_clock -> kops/s: 1604.85 (~= 24 * 1.25 * 53.944)
4181MB 64thread base -> kops/s: 307.839
4181MB 64thread folly -> kops/s: 272.172
4181MB 64thread gt_clock -> kops/s: 1204.47
4181MB 64thread new_clock -> kops/s: 1615.37
4181MB 128thread base -> kops/s: 310.934
4181MB 128thread folly -> kops/s: 267.468
4181MB 128thread gt_clock -> kops/s: 1188.75
4181MB 128thread new_clock -> kops/s: 1595.46
Whether we have just one thread on a quiet system or an overload of threads, the new version wins every time in thousand-ops per second, sometimes dramatically so. Mutex-based implementation quickly becomes contention-limited. New clock cache shows essentially perfect scaling up to number of physical cores (24), and then each hyperthreaded core adding about 1/4 the throughput of an additional physical core (see 48 thread case). Block cache miss rates (omitted above) are negligible across the board. With partitioned instead of full filters, the maximum speed-up vs. base is more like 2.5x rather than 5x.
Now test a large block cache with low miss ratio, but some eviction is required:
1597MB 1thread base -> kops/s: 46.603 io_bytes/op: 1584.63 miss_ratio: 0.0201066 max_rss_mb: 1589.23
1597MB 1thread folly -> kops/s: 45.079 io_bytes/op: 1530.03 miss_ratio: 0.019872 max_rss_mb: 1550.43
1597MB 1thread gt_clock -> kops/s: 48.711 io_bytes/op: 1566.63 miss_ratio: 0.0198923 max_rss_mb: 1691.4
1597MB 1thread new_clock -> kops/s: 51.531 io_bytes/op: 1589.07 miss_ratio: 0.0201969 max_rss_mb: 1583.56
1597MB 32thread base -> kops/s: 301.174 io_bytes/op: 1439.52 miss_ratio: 0.0184218 max_rss_mb: 1656.59
1597MB 32thread folly -> kops/s: 273.09 io_bytes/op: 1375.12 miss_ratio: 0.0180002 max_rss_mb: 1586.8
1597MB 32thread gt_clock -> kops/s: 904.497 io_bytes/op: 1411.29 miss_ratio: 0.0179934 max_rss_mb: 1775.89
1597MB 32thread new_clock -> kops/s: 1182.59 io_bytes/op: 1440.77 miss_ratio: 0.0185449 max_rss_mb: 1636.45
1597MB 128thread base -> kops/s: 309.91 io_bytes/op: 1438.25 miss_ratio: 0.018399 max_rss_mb: 1689.98
1597MB 128thread folly -> kops/s: 267.605 io_bytes/op: 1394.16 miss_ratio: 0.0180286 max_rss_mb: 1631.91
1597MB 128thread gt_clock -> kops/s: 691.518 io_bytes/op: 9056.73 miss_ratio: 0.0186572 max_rss_mb: 1982.26
1597MB 128thread new_clock -> kops/s: 1406.12 io_bytes/op: 1440.82 miss_ratio: 0.0185463 max_rss_mb: 1685.63
610MB 1thread base -> kops/s: 45.511 io_bytes/op: 2279.61 miss_ratio: 0.0290528 max_rss_mb: 615.137
610MB 1thread folly -> kops/s: 43.386 io_bytes/op: 2217.29 miss_ratio: 0.0289282 max_rss_mb: 600.996
610MB 1thread gt_clock -> kops/s: 46.207 io_bytes/op: 2275.51 miss_ratio: 0.0290057 max_rss_mb: 637.934
610MB 1thread new_clock -> kops/s: 48.879 io_bytes/op: 2283.1 miss_ratio: 0.0291253 max_rss_mb: 613.5
610MB 32thread base -> kops/s: 306.59 io_bytes/op: 2250 miss_ratio: 0.0288721 max_rss_mb: 683.402
610MB 32thread folly -> kops/s: 269.176 io_bytes/op: 2187.86 miss_ratio: 0.0286938 max_rss_mb: 628.742
610MB 32thread gt_clock -> kops/s: 855.097 io_bytes/op: 2279.26 miss_ratio: 0.0288009 max_rss_mb: 733.062
610MB 32thread new_clock -> kops/s: 1121.47 io_bytes/op: 2244.29 miss_ratio: 0.0289046 max_rss_mb: 666.453
610MB 128thread base -> kops/s: 305.079 io_bytes/op: 2252.43 miss_ratio: 0.0288884 max_rss_mb: 723.457
610MB 128thread folly -> kops/s: 269.583 io_bytes/op: 2204.58 miss_ratio: 0.0287001 max_rss_mb: 676.426
610MB 128thread gt_clock -> kops/s: 53.298 io_bytes/op: 8128.98 miss_ratio: 0.0292452 max_rss_mb: 956.273
610MB 128thread new_clock -> kops/s: 1301.09 io_bytes/op: 2246.04 miss_ratio: 0.0289171 max_rss_mb: 788.812
The new version is still winning every time, sometimes dramatically so, and we can tell from the maximum resident memory numbers (which contain some noise, by the way) that the new cache is not cheating on memory usage. IMPORTANT: The previous generation experimental clock cache appears to hit a serious bottleneck in the higher thread count configurations, presumably due to some of its waiting functionality. (The same bottleneck is not seen with partitioned index+filters.)
Now we consider even smaller cache sizes, with higher miss ratios, eviction work, etc.
233MB 1thread base -> kops/s: 10.557 io_bytes/op: 227040 miss_ratio: 0.0403105 max_rss_mb: 247.371
233MB 1thread folly -> kops/s: 15.348 io_bytes/op: 112007 miss_ratio: 0.0372238 max_rss_mb: 245.293
233MB 1thread gt_clock -> kops/s: 6.365 io_bytes/op: 244854 miss_ratio: 0.0413873 max_rss_mb: 259.844
233MB 1thread new_clock -> kops/s: 47.501 io_bytes/op: 2591.93 miss_ratio: 0.0330989 max_rss_mb: 242.461
233MB 32thread base -> kops/s: 96.498 io_bytes/op: 363379 miss_ratio: 0.0459966 max_rss_mb: 479.227
233MB 32thread folly -> kops/s: 109.95 io_bytes/op: 314799 miss_ratio: 0.0450032 max_rss_mb: 400.738
233MB 32thread gt_clock -> kops/s: 2.353 io_bytes/op: 385397 miss_ratio: 0.048445 max_rss_mb: 500.688
233MB 32thread new_clock -> kops/s: 1088.95 io_bytes/op: 2567.02 miss_ratio: 0.0330593 max_rss_mb: 303.402
233MB 128thread base -> kops/s: 84.302 io_bytes/op: 378020 miss_ratio: 0.0466558 max_rss_mb: 1051.84
233MB 128thread folly -> kops/s: 89.921 io_bytes/op: 338242 miss_ratio: 0.0460309 max_rss_mb: 812.785
233MB 128thread gt_clock -> kops/s: 2.588 io_bytes/op: 462833 miss_ratio: 0.0509158 max_rss_mb: 1109.94
233MB 128thread new_clock -> kops/s: 1299.26 io_bytes/op: 2565.94 miss_ratio: 0.0330531 max_rss_mb: 361.016
89MB 1thread base -> kops/s: 0.574 io_bytes/op: 5.35977e+06 miss_ratio: 0.274427 max_rss_mb: 91.3086
89MB 1thread folly -> kops/s: 0.578 io_bytes/op: 5.16549e+06 miss_ratio: 0.27276 max_rss_mb: 96.8984
89MB 1thread gt_clock -> kops/s: 0.512 io_bytes/op: 4.13111e+06 miss_ratio: 0.242817 max_rss_mb: 119.441
89MB 1thread new_clock -> kops/s: 48.172 io_bytes/op: 2709.76 miss_ratio: 0.0346162 max_rss_mb: 100.754
89MB 32thread base -> kops/s: 5.779 io_bytes/op: 6.14192e+06 miss_ratio: 0.320399 max_rss_mb: 311.812
89MB 32thread folly -> kops/s: 5.601 io_bytes/op: 5.83838e+06 miss_ratio: 0.313123 max_rss_mb: 252.418
89MB 32thread gt_clock -> kops/s: 0.77 io_bytes/op: 3.99236e+06 miss_ratio: 0.236296 max_rss_mb: 396.422
89MB 32thread new_clock -> kops/s: 1064.97 io_bytes/op: 2687.23 miss_ratio: 0.0346134 max_rss_mb: 155.293
89MB 128thread base -> kops/s: 4.959 io_bytes/op: 6.20297e+06 miss_ratio: 0.323945 max_rss_mb: 823.43
89MB 128thread folly -> kops/s: 4.962 io_bytes/op: 5.9601e+06 miss_ratio: 0.319857 max_rss_mb: 626.824
89MB 128thread gt_clock -> kops/s: 1.009 io_bytes/op: 4.1083e+06 miss_ratio: 0.242512 max_rss_mb: 1095.32
89MB 128thread new_clock -> kops/s: 1224.39 io_bytes/op: 2688.2 miss_ratio: 0.0346207 max_rss_mb: 218.223
^ Now something interesting has happened: the new clock cache has gained a dramatic lead in the single-threaded case, and this is because the cache is so small, and full filters are so big, that dividing the cache into 64 shards leads to significant (random) imbalances in cache shards and excessive churn in imbalanced shards. This new clock cache only uses two shards for this configuration, and that helps to ensure that entries are part of a sufficiently big pool that their eviction order resembles the single-shard order. (This effect is not seen with partitioned index+filters.)
Even smaller cache size:
34MB 1thread base -> kops/s: 0.198 io_bytes/op: 1.65342e+07 miss_ratio: 0.939466 max_rss_mb: 48.6914
34MB 1thread folly -> kops/s: 0.201 io_bytes/op: 1.63416e+07 miss_ratio: 0.939081 max_rss_mb: 45.3281
34MB 1thread gt_clock -> kops/s: 0.448 io_bytes/op: 4.43957e+06 miss_ratio: 0.266749 max_rss_mb: 100.523
34MB 1thread new_clock -> kops/s: 1.055 io_bytes/op: 1.85439e+06 miss_ratio: 0.107512 max_rss_mb: 75.3125
34MB 32thread base -> kops/s: 3.346 io_bytes/op: 1.64852e+07 miss_ratio: 0.93596 max_rss_mb: 180.48
34MB 32thread folly -> kops/s: 3.431 io_bytes/op: 1.62857e+07 miss_ratio: 0.935693 max_rss_mb: 137.531
34MB 32thread gt_clock -> kops/s: 1.47 io_bytes/op: 4.89704e+06 miss_ratio: 0.295081 max_rss_mb: 392.465
34MB 32thread new_clock -> kops/s: 8.19 io_bytes/op: 3.70456e+06 miss_ratio: 0.20826 max_rss_mb: 519.793
34MB 128thread base -> kops/s: 2.293 io_bytes/op: 1.64351e+07 miss_ratio: 0.931866 max_rss_mb: 449.484
34MB 128thread folly -> kops/s: 2.34 io_bytes/op: 1.6219e+07 miss_ratio: 0.932023 max_rss_mb: 396.457
34MB 128thread gt_clock -> kops/s: 1.798 io_bytes/op: 5.4241e+06 miss_ratio: 0.324881 max_rss_mb: 1104.41
34MB 128thread new_clock -> kops/s: 10.519 io_bytes/op: 2.39354e+06 miss_ratio: 0.136147 max_rss_mb: 1050.52
As the miss ratio gets higher (say, above 10%), the CPU time spent in eviction starts to erode the advantage of using fewer shards (13% miss rate much lower than 94%). LRU's O(1) eviction time can eventually pay off when there's enough block cache churn:
13MB 1thread base -> kops/s: 0.195 io_bytes/op: 1.65732e+07 miss_ratio: 0.946604 max_rss_mb: 45.6328
13MB 1thread folly -> kops/s: 0.197 io_bytes/op: 1.63793e+07 miss_ratio: 0.94661 max_rss_mb: 33.8633
13MB 1thread gt_clock -> kops/s: 0.519 io_bytes/op: 4.43316e+06 miss_ratio: 0.269379 max_rss_mb: 100.684
13MB 1thread new_clock -> kops/s: 0.176 io_bytes/op: 1.54148e+07 miss_ratio: 0.91545 max_rss_mb: 66.2383
13MB 32thread base -> kops/s: 3.266 io_bytes/op: 1.65544e+07 miss_ratio: 0.943386 max_rss_mb: 132.492
13MB 32thread folly -> kops/s: 3.396 io_bytes/op: 1.63142e+07 miss_ratio: 0.943243 max_rss_mb: 101.863
13MB 32thread gt_clock -> kops/s: 2.758 io_bytes/op: 5.13714e+06 miss_ratio: 0.310652 max_rss_mb: 396.121
13MB 32thread new_clock -> kops/s: 3.11 io_bytes/op: 1.23419e+07 miss_ratio: 0.708425 max_rss_mb: 321.758
13MB 128thread base -> kops/s: 2.31 io_bytes/op: 1.64823e+07 miss_ratio: 0.939543 max_rss_mb: 425.539
13MB 128thread folly -> kops/s: 2.339 io_bytes/op: 1.6242e+07 miss_ratio: 0.939966 max_rss_mb: 346.098
13MB 128thread gt_clock -> kops/s: 3.223 io_bytes/op: 5.76928e+06 miss_ratio: 0.345899 max_rss_mb: 1087.77
13MB 128thread new_clock -> kops/s: 2.984 io_bytes/op: 1.05341e+07 miss_ratio: 0.606198 max_rss_mb: 898.27
gt_clock is clearly blowing way past its memory budget for lower miss rates and best throughput. new_clock also seems to be exceeding budgets, and this warrants more investigation but is not the use case we are targeting with the new cache. With partitioned index+filter, the miss ratio is much better, and although still high enough that the eviction CPU time is definitely offsetting mutex contention:
13MB 1thread base -> kops/s: 16.326 io_bytes/op: 23743.9 miss_ratio: 0.205362 max_rss_mb: 65.2852
13MB 1thread folly -> kops/s: 15.574 io_bytes/op: 19415 miss_ratio: 0.184157 max_rss_mb: 56.3516
13MB 1thread gt_clock -> kops/s: 14.459 io_bytes/op: 22873 miss_ratio: 0.198355 max_rss_mb: 63.9688
13MB 1thread new_clock -> kops/s: 16.34 io_bytes/op: 24386.5 miss_ratio: 0.210512 max_rss_mb: 61.707
13MB 128thread base -> kops/s: 289.786 io_bytes/op: 23710.9 miss_ratio: 0.205056 max_rss_mb: 103.57
13MB 128thread folly -> kops/s: 185.282 io_bytes/op: 19433.1 miss_ratio: 0.184275 max_rss_mb: 116.219
13MB 128thread gt_clock -> kops/s: 354.451 io_bytes/op: 23150.6 miss_ratio: 0.200495 max_rss_mb: 102.871
13MB 128thread new_clock -> kops/s: 295.359 io_bytes/op: 24626.4 miss_ratio: 0.212452 max_rss_mb: 121.109
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10626
Test Plan: updated unit tests, stress/crash test runs including with TSAN, ASAN, UBSAN
Reviewed By: anand1976
Differential Revision: D39368406
Pulled By: pdillinger
fbshipit-source-id: 5afc44da4c656f8f751b44552bbf27bd3ca6fef9
Summary:
**Summary:**
When a block is firstly `Lookup` from the secondary cache, we just insert a dummy block in the primary cache (charging the actual size of the block) and don’t erase the block from the secondary cache. A standalone handle is returned from `Lookup`. Only if the block is hit again, we erase it from the secondary cache and add it into the primary cache.
When a block is firstly evicted from the primary cache to the secondary cache, we just insert a dummy block (size 0) in the secondary cache. When the block is evicted again, it is treated as a hot block and is inserted into the secondary cache.
**Implementation Details**
Add a new state of LRUHandle: The handle is never inserted into the LRUCache (both hash table and LRU list) and it doesn't experience the above three states. The entry can be freed when refs becomes 0. (refs >= 1 && in_cache == false && IS_STANDALONE == true)
The behaviors of `LRUCacheShard::Lookup()` are updated if the secondary_cache is CompressedSecondaryCache:
1. If a handle is found in primary cache:
1.1. If the handle's value is not nullptr, it is returned immediately.
1.2. If the handle's value is nullptr, this means the handle is a dummy one. For a dummy handle, if it was retrieved from secondary cache, it may still exist in secondary cache.
- 1.2.1. If no valid handle can be `Lookup` from secondary cache, return nullptr.
- 1.2.2. If the handle from secondary cache is valid, erase it from the secondary cache and add it into the primary cache.
2. If a handle is not found in primary cache:
2.1. If no valid handle can be `Lookup` from secondary cache, return nullptr.
2.2. If the handle from secondary cache is valid, insert a dummy block in the primary cache (charging the actual size of the block) and return a standalone handle.
The behaviors of `LRUCacheShard::Promote()` are updated as follows:
1. If `e->sec_handle` has value, one of the following steps can happen:
1.1. Insert a dummy handle and return a standalone handle to caller when `secondary_cache_` is `CompressedSecondaryCache` and e is a standalone handle.
1.2. Insert the item into the primary cache and return the handle to caller.
1.3. Exception handling.
3. If `e->sec_handle` has no value, mark the item as not in cache and charge the cache as its only metadata that'll shortly be released.
The behavior of `CompressedSecondaryCache::Insert()` is updated:
1. If a block is evicted from the primary cache for the first time, a dummy item is inserted.
4. If a dummy item is found for a block, the block is inserted into the secondary cache.
The behavior of `CompressedSecondaryCache:::Lookup()` is updated:
1. If a handle is not found or it is a dummy item, a nullptr is returned.
2. If `erase_handle` is true, the handle is erased.
The behaviors of `LRUCacheShard::Release()` are adjusted for the standalone handles.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10527
Test Plan:
1. stress tests.
5. unit tests.
6. CPU profiling for db_bench.
Reviewed By: siying
Differential Revision: D38747613
Pulled By: gitbw95
fbshipit-source-id: 74a1eba7e1957c9affb2bd2ae3e0194584fa6eca
Summary:
RocksDB's `Cache` abstraction currently supports two priority levels for items: high (used for frequently accessed/highly valuable SST metablocks like index/filter blocks) and low (used for SST data blocks). Blobs are typically lower-value targets for caching than data blocks, since 1) with BlobDB, data blocks containing blob references conceptually form an index structure which has to be consulted before we can read the blob value, and 2) cached blobs represent only a single key-value, while cached data blocks generally contain multiple KVs. Since we would like to make it possible to use the same backing cache for the block cache and the blob cache, it would make sense to add a new, lower-than-low cache priority level (bottom level) for blobs so data blocks are prioritized over them.
This task is a part of https://github.com/facebook/rocksdb/issues/10156
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10461
Reviewed By: siying
Differential Revision: D38672823
Pulled By: ltamasi
fbshipit-source-id: 90cf7362036563d79891f47be2cc24b827482743
Summary:
Currently, `SetIsInSecondaryCache` is after `Promote`. After `Promote`, a handle can be accessed and its flags can be set. This causes data race.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10472
Test Plan:
unit tests
stress tests
Reviewed By: pdillinger
Differential Revision: D38403991
Pulled By: gitbw95
fbshipit-source-id: 0aaa2d2edeaf5bc799fcce605648fe49eb7119c2
Summary:
If a secondary cache is configured, its possible that a cache lookup will get a hit in the secondary cache. In that case, the ```LRUCacheShard::Lookup``` doesn't immediately update the ```total_charge``` for the item handle if the ```wait``` parameter is false (i.e caller will call later to check the completeness). However, ```BlockBasedTable::GetEntryFromCache``` assumes the handle is complete and calls ```UpdateCacheHitMetrics```, which checks the usage of the cache item and fails the assert in https://github.com/facebook/rocksdb/blob/main/cache/lru_cache.h#L237 (```assert(total_charge >= meta_charge)```).
To fix this, we call ```UpdateCacheHitMetrics``` later in ```MultiGet```, after waiting for all cache lookup completions.
Test plan -
Run crash test with changes from https://github.com/facebook/rocksdb/issues/10160
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10440
Reviewed By: gitbw95
Differential Revision: D38283968
Pulled By: anand1976
fbshipit-source-id: 31c54ef43517726c6e5fdda81899b364241dd7e1
Summary:
RocksDB's `Cache` abstraction currently supports two priority levels for items: high (used for frequently accessed/highly valuable SST metablocks like index/filter blocks) and low (used for SST data blocks). Blobs are typically lower-value targets for caching than data blocks, since 1) with BlobDB, data blocks containing blob references conceptually form an index structure which has to be consulted before we can read the blob value, and 2) cached blobs represent only a single key-value, while cached data blocks generally contain multiple KVs. Since we would like to make it possible to use the same backing cache for the block cache and the blob cache, it would make sense to add a new, lower-than-low cache priority level (bottom level) for blobs so data blocks are prioritized over them.
This task is a part of https://github.com/facebook/rocksdb/issues/10156
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10309
Reviewed By: ltamasi
Differential Revision: D38211655
Pulled By: gangliao
fbshipit-source-id: 65ef33337db4d85277cc6f9782d67c421ad71dd5
Summary:
RocksDB supports a two-level cache hierarchy (see https://rocksdb.org/blog/2021/05/27/rocksdb-secondary-cache.html), where items evicted from the primary cache can be spilled over to the secondary cache, or items from the secondary cache can be promoted to the primary one. We have a CacheLib-based non-volatile secondary cache implementation that can be used to improve read latencies and reduce the amount of network bandwidth when using distributed file systems. In addition, we have recently implemented a compressed secondary cache that can be used as a replacement for the OS page cache when e.g. direct I/O is used. The goals of this task are to add support for using a secondary cache with the blob cache and to measure the potential performance gains using `db_bench`.
This task is a part of https://github.com/facebook/rocksdb/issues/10156
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10349
Reviewed By: ltamasi
Differential Revision: D37896773
Pulled By: gangliao
fbshipit-source-id: 7804619ce4a44b73d9e11ad606640f9385969c84
Summary:
If the primary cache is LRU cache and there is a secondary cache, add Secondary Cache printable options into LRUCache::GetPrintableOptions.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10346
Test Plan:
1. Current Unit Tests should pass.
2. Use db_bench (with compressed_secondary_cache ) and the LOG should includes the new printable options from Seoncdary Cache.
Reviewed By: jay-zhuang
Differential Revision: D37779310
Pulled By: gitbw95
fbshipit-source-id: 88ce1f7df6b5f25740e598d9e7fa91e4c414cb8f
Summary:
I noticed it would clean up some things to have Cache::Insert()
return our MemoryLimit Status instead of Incomplete for the case in
which the capacity limit is reached. I suspect this fixes some existing but
unknown bugs where this Incomplete could be confused with other uses
of Incomplete, especially no_io cases. This is the most suspicious case I
noticed, but was not able to reproduce a bug, in part because the existing
code is not covered by unit tests (FIXME added): 57adbf0e91/table/get_context.cc (L397)
I audited all the existing uses of IsIncomplete and updated those that
seemed relevant.
HISTORY updated with a clear warning to users of strict_capacity_limit=true
to update uses of `IsIncomplete()`
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10262
Test Plan: updated unit tests
Reviewed By: hx235
Differential Revision: D37473155
Pulled By: pdillinger
fbshipit-source-id: 4bd9d9353ccddfe286b03ebd0652df8ce20f99cb
Summary:
folly DistributedMutex is faster than standard mutexes though
imposes some static obligations on usage. See
https://github.com/facebook/folly/blob/main/folly/synchronization/DistributedMutex.h
for details. Here we use this alternative for our Cache implementations
(especially LRUCache) for better locking performance, when RocksDB is
compiled with folly.
Also added information about which distributed mutex implementation is
being used to cache_bench output and to DB LOG.
Intended follow-up:
* Use DMutex in more places, perhaps improving API to support non-scoped
locking
* Fix linking with fbcode compiler (needs ROCKSDB_NO_FBCODE=1 currently)
Credit: Thanks Siying for reminding me about this line of work that was previously
left unfinished.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10179
Test Plan:
for correctness, existing tests. CircleCI config updated.
Also Meta-internal buck build updated.
For performance, ran simultaneous before & after cache_bench. Out of three
comparison runs, the middle improvement to ops/sec was +21%:
Baseline: USE_CLANG=1 DEBUG_LEVEL=0 make -j24 cache_bench (fbcode
compiler)
```
Complete in 20.201 s; Rough parallel ops/sec = 1584062
Thread ops/sec = 107176
Operation latency (ns):
Count: 32000000 Average: 9257.9421 StdDev: 122412.04
Min: 134 Median: 3623.0493 Max: 56918500
Percentiles: P50: 3623.05 P75: 10288.02 P99: 30219.35 P99.9: 683522.04 P99.99: 7302791.63
```
New: (add USE_FOLLY=1)
```
Complete in 16.674 s; Rough parallel ops/sec = 1919135 (+21%)
Thread ops/sec = 135487
Operation latency (ns):
Count: 32000000 Average: 7304.9294 StdDev: 108530.28
Min: 132 Median: 3777.6012 Max: 91030902
Percentiles: P50: 3777.60 P75: 10169.89 P99: 24504.51 P99.9: 59721.59 P99.99: 1861151.83
```
Reviewed By: anand1976
Differential Revision: D37182983
Pulled By: pdillinger
fbshipit-source-id: a17eb05f25b832b6a2c1356f5c657e831a5af8d1
Summary:
In LRU Cache mutex, we sometimes call malloc_usable_size() to calculate memory used by the metadata object. We prevent it by saving the charge + metadata size, rather than charge, inside the metadata itself. Within the mutex, usually only total charge is needed so we don't need to repeat.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10026
Test Plan: Run existing tests.
Reviewed By: pdillinger
Differential Revision: D36556253
fbshipit-source-id: f60c96d13cde3af77732e5548e4eac4182fa9801
Summary:
To support a project to prototype and evaluate algorithmic
enhancments and alternatives to LRUCache, here I have separated out
LRUCache into internal-only "FastLRUCache" and cut it down to
essentials, so that details like secondary cache handling and
priorities do not interfere with prototyping. These can be
re-integrated later as needed, along with refactoring to minimize code
duplication (which would slow down prototyping for now).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9917
Test Plan:
unit tests updated to ensure basic functionality has (likely)
been preserved
Reviewed By: anand1976
Differential Revision: D35995554
Pulled By: pdillinger
fbshipit-source-id: d67b20b7ada3b5d3bfe56d897a73885894a1d9db
Summary:
The param name force_erase may be misleading, since the handle is erased only if it has last reference even if the param is set true.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9728
Reviewed By: pdillinger
Differential Revision: D35038673
Pulled By: gitbw95
fbshipit-source-id: 0d16d1e8fed17b97eba7fb53207119332f659a5f
Summary:
Added missing include, and cleaned up to make same mistake less
likely in future (minimize conditional compilation)
Fixes https://github.com/facebook/rocksdb/issues/9183
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9209
Test Plan: added to existing test
Reviewed By: mrambacher
Differential Revision: D32631390
Pulled By: pdillinger
fbshipit-source-id: 63a0501855cf5fac9e22ca1e5c4f53725dbf3f93
Summary:
In ```LRUCacheShard::Promote```, a reference is released outside the LRU mutex. Fix the race condition.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8717
Reviewed By: zhichao-cao
Differential Revision: D30649206
Pulled By: anand1976
fbshipit-source-id: 09c0af05b2294a7fe2c02876a61b0bad6e3ada61
Summary:
Add a stat for secondary cache hits. The ```Cache::Lookup``` API had an unused ```stats``` parameter. This PR uses that to pass the pointer to a ```Statistics``` object that ```LRUCache``` uses to record the stat.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8666
Test Plan: Update a unit test in lru_cache_test
Reviewed By: zhichao-cao
Differential Revision: D30353816
Pulled By: anand1976
fbshipit-source-id: 2046f78b460428877a26ffdd2bb914ae47dfbe77
Summary:
Some bits are mutated and read while holding a lock, other
immutable bits (esp. secondary cache compatibility) can be read by
arbitrary threads without holding a lock. AFAIK, this doesn't cause an
issue on any architecture we care about, because you will get some
legitimate version of the value that includes the initialization, as
long as synchronization guarantees the initialization happens before the
read.
I've only seen this in https://github.com/facebook/rocksdb/issues/8538 so far, but it should be fixed regardless.
Otherwise, we'll surely get these false reports again some time.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8539
Test Plan: some local TSAN test runs and in CircleCI
Reviewed By: zhichao-cao
Differential Revision: D29720262
Pulled By: pdillinger
fbshipit-source-id: 365fd7e565577c648815161f71b339bcb5ce12d5
Summary:
When the secondary cache lookup fails, we may still allocate a handle and charge the cache for metadata usage. If the cache is full, this can cause the usage to go over capacity. Later, when a (unrelated) handle is released, it trips up an assertion that checks that usage is less than capacity. To prevent this assertion failure, don't charge the cache for a failed secondary cache lookup.
Tests:
Run crash_test
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8470
Reviewed By: zhichao-cao
Differential Revision: D29474713
Pulled By: anand1976
fbshipit-source-id: 27191969c95470a7b070d292b458efce71395bf2
Summary:
Tsan complains due to a perceived race condition in accessing LRUHandle flags. One thread calls ```LRUHandle::SetHit()``` from ```LRUCacheShard::Lookup()```, while another thread calls ```LRUHandle::IsPending()``` from ```LRUCacheShard::IsReady()```. The latter call is from ```MultiGet```. It doesn't actually have to call ```IsReady``` since a null value indicates the cache handle is not ready, so its sufficient to check for a null value.
Also modify ```IsReady``` to acquire the LRU shard mutex.
Tests:
1. make check
2. Run tsan_crash
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8433
Reviewed By: zhichao-cao
Differential Revision: D29278030
Pulled By: anand1976
fbshipit-source-id: 0c9fed56d12eda853e72dadebe75038361bd257f
Summary:
Implement the ```WaitAll()``` interface in ```LRUCache``` to allow callers to issue multiple lookups in parallel and wait for all of them to complete. Modify ```MultiGet``` to use this to parallelize the secondary cache lookups in order to reduce the overall latency. A call to ```cache->Lookup()``` returns a handle that has an incomplete value (nullptr), and the caller can call ```cache->IsReady()``` to check whether the lookup is complete, and pass a vector of handles to ```WaitAll``` to wait for completion. If any of the lookups fail, ```MultiGet``` will read the block from the SST file.
Another change in this PR is to rename ```SecondaryCacheHandle``` to ```SecondaryCacheResultHandle``` as it more accurately describes the return result of the secondary cache lookup, which is more like a future.
Tests:
1. Add unit tests in lru_cache_test
2. Benchmark results with no secondary cache configured
Master -
```
readrandom : 41.175 micros/op 388562 ops/sec; 106.7 MB/s (7277999 of 7277999 found)
readrandom : 41.217 micros/op 388160 ops/sec; 106.6 MB/s (7274999 of 7274999 found)
multireadrandom : 10.309 micros/op 1552082 ops/sec; (28908992 of 28908992 found)
multireadrandom : 10.321 micros/op 1550218 ops/sec; (29081984 of 29081984 found)
```
This PR -
```
readrandom : 41.158 micros/op 388723 ops/sec; 106.8 MB/s (7290999 of 7290999 found)
readrandom : 41.185 micros/op 388463 ops/sec; 106.7 MB/s (7287999 of 7287999 found)
multireadrandom : 10.277 micros/op 1556801 ops/sec; (29346944 of 29346944 found)
multireadrandom : 10.253 micros/op 1560539 ops/sec; (29274944 of 29274944 found)
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8405
Reviewed By: zhichao-cao
Differential Revision: D29190509
Pulled By: anand1976
fbshipit-source-id: 6f8eff6246712af8a297cfe22ea0d1c3b2a01bb0
Summary:
This change gathers and publishes statistics about the
kinds of items in block cache. This is especially important for
profiling relative usage of cache by index vs. filter vs. data blocks.
It works by iterating over the cache during periodic stats dump
(InternalStats, stats_dump_period_sec) or on demand when
DB::Get(Map)Property(kBlockCacheEntryStats), except that for
efficiency and sharing among column families, saved data from
the last scan is used when the data is not considered too old.
The new information can be seen in info LOG, for example:
Block cache LRUCache@0x7fca62229330 capacity: 95.37 MB collections: 8 last_copies: 0 last_secs: 0.00178 secs_since: 0
Block cache entry stats(count,size,portion): DataBlock(7092,28.24 MB,29.6136%) FilterBlock(215,867.90 KB,0.888728%) FilterMetaBlock(2,5.31 KB,0.00544%) IndexBlock(217,180.11 KB,0.184432%) WriteBuffer(1,256.00 KB,0.262144%) Misc(1,0.00 KB,0%)
And also through DB::GetProperty and GetMapProperty (here using
ldb just for demonstration):
$ ./ldb --db=/dev/shm/dbbench/ get_property rocksdb.block-cache-entry-stats
rocksdb.block-cache-entry-stats.bytes.data-block: 0
rocksdb.block-cache-entry-stats.bytes.deprecated-filter-block: 0
rocksdb.block-cache-entry-stats.bytes.filter-block: 0
rocksdb.block-cache-entry-stats.bytes.filter-meta-block: 0
rocksdb.block-cache-entry-stats.bytes.index-block: 178992
rocksdb.block-cache-entry-stats.bytes.misc: 0
rocksdb.block-cache-entry-stats.bytes.other-block: 0
rocksdb.block-cache-entry-stats.bytes.write-buffer: 0
rocksdb.block-cache-entry-stats.capacity: 8388608
rocksdb.block-cache-entry-stats.count.data-block: 0
rocksdb.block-cache-entry-stats.count.deprecated-filter-block: 0
rocksdb.block-cache-entry-stats.count.filter-block: 0
rocksdb.block-cache-entry-stats.count.filter-meta-block: 0
rocksdb.block-cache-entry-stats.count.index-block: 215
rocksdb.block-cache-entry-stats.count.misc: 1
rocksdb.block-cache-entry-stats.count.other-block: 0
rocksdb.block-cache-entry-stats.count.write-buffer: 0
rocksdb.block-cache-entry-stats.id: LRUCache@0x7f3636661290
rocksdb.block-cache-entry-stats.percent.data-block: 0.000000
rocksdb.block-cache-entry-stats.percent.deprecated-filter-block: 0.000000
rocksdb.block-cache-entry-stats.percent.filter-block: 0.000000
rocksdb.block-cache-entry-stats.percent.filter-meta-block: 0.000000
rocksdb.block-cache-entry-stats.percent.index-block: 2.133751
rocksdb.block-cache-entry-stats.percent.misc: 0.000000
rocksdb.block-cache-entry-stats.percent.other-block: 0.000000
rocksdb.block-cache-entry-stats.percent.write-buffer: 0.000000
rocksdb.block-cache-entry-stats.secs_for_last_collection: 0.000052
rocksdb.block-cache-entry-stats.secs_since_last_collection: 0
Solution detail - We need some way to flag what kind of blocks each
entry belongs to, preferably without changing the Cache API.
One of the complications is that Cache is a general interface that could
have other users that don't adhere to whichever convention we decide
on for keys and values. Or we would pay for an extra field in the Handle
that would only be used for this purpose.
This change uses a back-door approach, the deleter, to indicate the
"role" of a Cache entry (in addition to the value type, implicitly).
This has the added benefit of ensuring proper code origin whenever we
recognize a particular role for a cache entry; if the entry came from
some other part of the code, it will use an unrecognized deleter, which
we simply attribute to the "Misc" role.
An internal API makes for simple instantiation and automatic
registration of Cache deleters for a given value type and "role".
Another internal API, CacheEntryStatsCollector, solves the problem of
caching the results of a scan and sharing them, to ensure scans are
neither excessive nor redundant so as not to harm Cache performance.
Because code is added to BlocklikeTraits, it is pulled out of
block_based_table_reader.cc into its own file.
This is a reformulation of https://github.com/facebook/rocksdb/issues/8276, without the type checking option
(could still be added), and with actual stat gathering.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8297
Test Plan: manual testing with db_bench, and a couple of basic unit tests
Reviewed By: ltamasi
Differential Revision: D28488721
Pulled By: pdillinger
fbshipit-source-id: 472f524a9691b5afb107934be2d41d84f2b129fb
Summary:
Defined the abstract interface for a secondary cache in include/rocksdb/secondary_cache.h, and updated LRUCacheOptions to take a std::shared_ptr<SecondaryCache>. An item is initially inserted into the LRU (primary) cache. When it ages out and evicted from memory, its inserted into the secondary cache. On a LRU cache miss and successful lookup in the secondary cache, the item is promoted to the LRU cache. Only support synchronous lookup currently. The secondary cache would be used to implement a persistent (flash cache) or compressed cache.
Tests:
Results from cache_bench and db_bench don't show any regression due to these changes.
cache_bench results before and after this change -
Command
```./cache_bench -ops_per_thread=10000000 -threads=1```
Before
```Complete in 40.688 s; QPS = 245774```
```Complete in 40.486 s; QPS = 246996```
```Complete in 42.019 s; QPS = 237989```
After
```Complete in 40.672 s; QPS = 245869```
```Complete in 44.622 s; QPS = 224107```
```Complete in 42.445 s; QPS = 235599```
db_bench results before this change, and with this change + https://github.com/facebook/rocksdb/issues/8213 and https://github.com/facebook/rocksdb/issues/8191 -
Commands
```./db_bench --benchmarks="fillseq,compact" -num=30000000 -key_size=32 -value_size=256 -use_direct_io_for_flush_and_compaction=true -db=/home/anand76/nvm_cache/db -partition_index_and_filters=true```
```./db_bench -db=/home/anand76/nvm_cache/db -use_existing_db=true -benchmarks=readrandom -num=30000000 -key_size=32 -value_size=256 -use_direct_reads=true -cache_size=1073741824 -cache_numshardbits=6 -cache_index_and_filter_blocks=true -read_random_exp_range=17 -statistics -partition_index_and_filters=true -threads=16 -duration=300```
Before
```
DB path: [/home/anand76/nvm_cache/db]
readrandom : 80.702 micros/op 198104 ops/sec; 54.4 MB/s (3708999 of 3708999 found)
```
```
DB path: [/home/anand76/nvm_cache/db]
readrandom : 87.124 micros/op 183625 ops/sec; 50.4 MB/s (3439999 of 3439999 found)
```
After
```
DB path: [/home/anand76/nvm_cache/db]
readrandom : 77.653 micros/op 206025 ops/sec; 56.6 MB/s (3866999 of 3866999 found)
```
```
DB path: [/home/anand76/nvm_cache/db]
readrandom : 84.962 micros/op 188299 ops/sec; 51.7 MB/s (3535999 of 3535999 found)
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8271
Reviewed By: zhichao-cao
Differential Revision: D28357511
Pulled By: anand1976
fbshipit-source-id: d1cfa236f00e649a18c53328be10a8062a4b6da2
Summary:
Adds a new Cache::ApplyToAllEntries API that we expect to use
(in follow-up PRs) for efficiently gathering block cache statistics.
Notable features vs. old ApplyToAllCacheEntries:
* Includes key and deleter (in addition to value and charge). We could
have passed in a Handle but then more virtual function calls would be
needed to get the "fields" of each entry. We expect to use the 'deleter'
to identify the origin of entries, perhaps even more.
* Heavily tuned to minimize latency impact on operating cache. It
does this by iterating over small sections of each cache shard while
cycling through the shards.
* Supports tuning roughly how many entries to operate on for each
lock acquire and release, to control the impact on the latency of other
operations without excessive lock acquire & release. The right balance
can depend on the cost of the callback. Good default seems to be
around 256.
* There should be no need to disable thread safety. (I would expect
uncontended locks to be sufficiently fast.)
I have enhanced cache_bench to validate this approach:
* Reports a histogram of ns per operation, so we can look at the
ditribution of times, not just throughput (average).
* Can add a thread for simulated "gather stats" which calls
ApplyToAllEntries at a specified interval. We also generate a histogram
of time to run ApplyToAllEntries.
To make the iteration over some entries of each shard work as cleanly as
possible, even with resize between next set of entries, I have
re-arranged which hash bits are used for sharding and which for indexing
within a shard.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8225
Test Plan:
A couple of unit tests are added, but primary validation is manual, as
the primary risk is to performance.
The primary validation is using cache_bench to ensure that neither
the minor hashing changes nor the simulated stats gathering
significantly impact QPS or latency distribution. Note that adding op
latency histogram seriously impacts the benchmark QPS, so for a
fair baseline, we need the cache_bench changes (except remove simulated
stat gathering to make it compile). In short, we don't see any
reproducible difference in ops/sec or op latency unless we are gathering
stats nearly continuously. Test uses 10GB block cache with
8KB values to be somewhat realistic in the number of items to iterate
over.
Baseline typical output:
```
Complete in 92.017 s; Rough parallel ops/sec = 869401
Thread ops/sec = 54662
Operation latency (ns):
Count: 80000000 Average: 11223.9494 StdDev: 29.61
Min: 0 Median: 7759.3973 Max: 9620500
Percentiles: P50: 7759.40 P75: 14190.73 P99: 46922.75 P99.9: 77509.84 P99.99: 217030.58
------------------------------------------------------
[ 0, 1 ] 68 0.000% 0.000%
( 2900, 4400 ] 89 0.000% 0.000%
( 4400, 6600 ] 33630240 42.038% 42.038% ########
( 6600, 9900 ] 18129842 22.662% 64.700% #####
( 9900, 14000 ] 7877533 9.847% 74.547% ##
( 14000, 22000 ] 15193238 18.992% 93.539% ####
( 22000, 33000 ] 3037061 3.796% 97.335% #
( 33000, 50000 ] 1626316 2.033% 99.368%
( 50000, 75000 ] 421532 0.527% 99.895%
( 75000, 110000 ] 56910 0.071% 99.966%
( 110000, 170000 ] 16134 0.020% 99.986%
( 170000, 250000 ] 5166 0.006% 99.993%
( 250000, 380000 ] 3017 0.004% 99.996%
( 380000, 570000 ] 1337 0.002% 99.998%
( 570000, 860000 ] 805 0.001% 99.999%
( 860000, 1200000 ] 319 0.000% 100.000%
( 1200000, 1900000 ] 231 0.000% 100.000%
( 1900000, 2900000 ] 100 0.000% 100.000%
( 2900000, 4300000 ] 39 0.000% 100.000%
( 4300000, 6500000 ] 16 0.000% 100.000%
( 6500000, 9800000 ] 7 0.000% 100.000%
```
New, gather_stats=false. Median thread ops/sec of 5 runs:
```
Complete in 92.030 s; Rough parallel ops/sec = 869285
Thread ops/sec = 54458
Operation latency (ns):
Count: 80000000 Average: 11298.1027 StdDev: 42.18
Min: 0 Median: 7722.0822 Max: 6398720
Percentiles: P50: 7722.08 P75: 14294.68 P99: 47522.95 P99.9: 85292.16 P99.99: 228077.78
------------------------------------------------------
[ 0, 1 ] 109 0.000% 0.000%
( 2900, 4400 ] 793 0.001% 0.001%
( 4400, 6600 ] 34054563 42.568% 42.569% #########
( 6600, 9900 ] 17482646 21.853% 64.423% ####
( 9900, 14000 ] 7908180 9.885% 74.308% ##
( 14000, 22000 ] 15032072 18.790% 93.098% ####
( 22000, 33000 ] 3237834 4.047% 97.145% #
( 33000, 50000 ] 1736882 2.171% 99.316%
( 50000, 75000 ] 446851 0.559% 99.875%
( 75000, 110000 ] 68251 0.085% 99.960%
( 110000, 170000 ] 18592 0.023% 99.983%
( 170000, 250000 ] 7200 0.009% 99.992%
( 250000, 380000 ] 3334 0.004% 99.997%
( 380000, 570000 ] 1393 0.002% 99.998%
( 570000, 860000 ] 700 0.001% 99.999%
( 860000, 1200000 ] 293 0.000% 100.000%
( 1200000, 1900000 ] 196 0.000% 100.000%
( 1900000, 2900000 ] 69 0.000% 100.000%
( 2900000, 4300000 ] 32 0.000% 100.000%
( 4300000, 6500000 ] 10 0.000% 100.000%
```
New, gather_stats=true, 1 second delay between scans. Scans take about
1 second here so it's spending about 50% time scanning. Still the effect on
ops/sec and latency seems to be in the noise. Median thread ops/sec of 5 runs:
```
Complete in 91.890 s; Rough parallel ops/sec = 870608
Thread ops/sec = 54551
Operation latency (ns):
Count: 80000000 Average: 11311.2629 StdDev: 45.28
Min: 0 Median: 7686.5458 Max: 10018340
Percentiles: P50: 7686.55 P75: 14481.95 P99: 47232.60 P99.9: 79230.18 P99.99: 232998.86
------------------------------------------------------
[ 0, 1 ] 71 0.000% 0.000%
( 2900, 4400 ] 291 0.000% 0.000%
( 4400, 6600 ] 34492060 43.115% 43.116% #########
( 6600, 9900 ] 16727328 20.909% 64.025% ####
( 9900, 14000 ] 7845828 9.807% 73.832% ##
( 14000, 22000 ] 15510654 19.388% 93.220% ####
( 22000, 33000 ] 3216533 4.021% 97.241% #
( 33000, 50000 ] 1680859 2.101% 99.342%
( 50000, 75000 ] 439059 0.549% 99.891%
( 75000, 110000 ] 60540 0.076% 99.967%
( 110000, 170000 ] 14649 0.018% 99.985%
( 170000, 250000 ] 5242 0.007% 99.991%
( 250000, 380000 ] 3260 0.004% 99.995%
( 380000, 570000 ] 1599 0.002% 99.997%
( 570000, 860000 ] 1043 0.001% 99.999%
( 860000, 1200000 ] 471 0.001% 99.999%
( 1200000, 1900000 ] 275 0.000% 100.000%
( 1900000, 2900000 ] 143 0.000% 100.000%
( 2900000, 4300000 ] 60 0.000% 100.000%
( 4300000, 6500000 ] 27 0.000% 100.000%
( 6500000, 9800000 ] 7 0.000% 100.000%
( 9800000, 14000000 ] 1 0.000% 100.000%
Gather stats latency (us):
Count: 46 Average: 980387.5870 StdDev: 60911.18
Min: 879155 Median: 1033777.7778 Max: 1261431
Percentiles: P50: 1033777.78 P75: 1120666.67 P99: 1261431.00 P99.9: 1261431.00 P99.99: 1261431.00
------------------------------------------------------
( 860000, 1200000 ] 45 97.826% 97.826% ####################
( 1200000, 1900000 ] 1 2.174% 100.000%
Most recent cache entry stats:
Number of entries: 1295133
Total charge: 9.88 GB
Average key size: 23.4982
Average charge: 8.00 KB
Unique deleters: 3
```
Reviewed By: mrambacher
Differential Revision: D28295742
Pulled By: pdillinger
fbshipit-source-id: bbc4a552f91ba0fe10e5cc025c42cef5a81f2b95
Summary:
This is a PR generated **semi-automatically** by an internal tool to remove unused includes and `using` statements.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7604
Test Plan: make check
Reviewed By: ajkr
Differential Revision: D24579392
Pulled By: riversand963
fbshipit-source-id: c4bfa6c6b08da1de186690d37eb73d8fff45aecd
Summary:
Since read threads do not coordinate on loading data into block
cache, two threads between Lookup and Insert can end up loading and
inserting the same data. This is particularly concerning with
cache_index_and_filter_blocks since those are hot and more likely to
be race targets if ejected from (or not pre-populated in) the cache.
Particularly with moves toward disaggregated / network storage, the cost
of redundant retrieval might be high, and we should at least have some
hard statistics from which we can estimate impact.
Example with full filter thrashing "cliff":
$ ./db_bench --benchmarks=fillrandom --num=15000000 --cache_index_and_filter_blocks -bloom_bits=10
...
$ ./db_bench --db=/tmp/rocksdbtest-172704/dbbench --use_existing_db --benchmarks=readrandom,stats --num=200000 --cache_index_and_filter_blocks --cache_size=$((130 * 1024 * 1024)) --bloom_bits=10 --threads=16 -statistics 2>&1 | egrep '^rocksdb.block.cache.(.*add|.*redundant)' | grep -v compress | sort
rocksdb.block.cache.add COUNT : 14181
rocksdb.block.cache.add.failures COUNT : 0
rocksdb.block.cache.add.redundant COUNT : 476
rocksdb.block.cache.data.add COUNT : 12749
rocksdb.block.cache.data.add.redundant COUNT : 18
rocksdb.block.cache.filter.add COUNT : 1003
rocksdb.block.cache.filter.add.redundant COUNT : 217
rocksdb.block.cache.index.add COUNT : 429
rocksdb.block.cache.index.add.redundant COUNT : 241
$ ./db_bench --db=/tmp/rocksdbtest-172704/dbbench --use_existing_db --benchmarks=readrandom,stats --num=200000 --cache_index_and_filter_blocks --cache_size=$((120 * 1024 * 1024)) --bloom_bits=10 --threads=16 -statistics 2>&1 | egrep '^rocksdb.block.cache.(.*add|.*redundant)' | grep -v compress | sort
rocksdb.block.cache.add COUNT : 1182223
rocksdb.block.cache.add.failures COUNT : 0
rocksdb.block.cache.add.redundant COUNT : 302728
rocksdb.block.cache.data.add COUNT : 31425
rocksdb.block.cache.data.add.redundant COUNT : 12
rocksdb.block.cache.filter.add COUNT : 795455
rocksdb.block.cache.filter.add.redundant COUNT : 130238
rocksdb.block.cache.index.add COUNT : 355343
rocksdb.block.cache.index.add.redundant COUNT : 172478
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6681
Test Plan: Some manual testing (above) and unit test covering key metrics is included
Reviewed By: ltamasi
Differential Revision: D21134113
Pulled By: pdillinger
fbshipit-source-id: c11497b5f00f4ffdfe919823904e52d0a1a91d87
Summary:
As the first step of reintroducing eviction statistics for the block
cache, the patch switches from using simple function pointers as deleters
to function objects implementing an interface. This will enable using
deleters that have state, like a smart pointer to the statistics object
that is to be updated when an entry is removed from the cache. For now,
the patch adds a deleter template class `SimpleDeleter`, which simply
casts the `value` pointer to its original type and calls `delete` or
`delete[]` on it as appropriate. Note: to prevent object lifecycle
issues, deleters must outlive the cache entries referring to them;
`SimpleDeleter` ensures this by using the ("leaky") Meyers singleton
pattern.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6545
Test Plan: `make asan_check`
Reviewed By: siying
Differential Revision: D20475823
Pulled By: ltamasi
fbshipit-source-id: fe354c33dd96d9bafc094605462352305449a22a
Summary:
When dynamically linking two binaries together, different builds of RocksDB from two sources might cause errors. To provide a tool for user to solve the problem, the RocksDB namespace is changed to a flag which can be overridden in build time.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6433
Test Plan: Build release, all and jtest. Try to build with ROCKSDB_NAMESPACE with another flag.
Differential Revision: D19977691
fbshipit-source-id: aa7f2d0972e1c31d75339ac48478f34f6cfcfb3e
Summary:
https://github.com/facebook/rocksdb/issues/5797 charges the block cache with the total of user-provided charge plus the metadata charge. It had a bug where in MaintainPoolSize the user-provided charge was used instead of the total charge. The patch fixes that.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5813
Differential Revision: D17412783
Pulled By: maysamyabandeh
fbshipit-source-id: 45c0ac9f1e2233760db5ccd61399605cd74edc87
Summary:
For our default block cache, each additional entry has extra memory overhead. It include LRUHandle (72 bytes currently) and the cache key (two varint64, file id and offset). The usage is not negligible. For example for block_size=4k, the overhead accounts for an extra 2% memory usage for the cache. The patch charging the cache for the extra usage, reducing untracked memory usage outside block cache. The feature is enabled by default and can be disabled by passing kDontChargeCacheMetadata to the cache constructor.
This PR builds up on https://github.com/facebook/rocksdb/issues/4258
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5797
Test Plan:
- Existing tests are updated to either disable the feature when the test has too much dependency on the old way of accounting the usage or increasing the cache capacity to account for the additional charge of metadata.
- The Usage tests in cache_test.cc are augmented to test the cache usage under kFullChargeCacheMetadata.
Differential Revision: D17396833
Pulled By: maysamyabandeh
fbshipit-source-id: 7684ccb9f8a40ca595e4f5efcdb03623afea0c6f
Summary:
The 'refs' field in LRUHandle now counts only external references, since anyway we already have the IN_CACHE flag. This simplifies reference accounting logic a bit. Also cleaned up few asserts code as well as the comments - to be more readable.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5579
Differential Revision: D16286747
Pulled By: elipoz
fbshipit-source-id: 7186d88f80f512ce584d0a303437494b5cbefd7f
Summary:
When using `PRIu64` type of printf specifier, current code base does the following:
```
#ifndef __STDC_FORMAT_MACROS
#define __STDC_FORMAT_MACROS
#endif
#include <inttypes.h>
```
However, this can be simplified to
```
#include <cinttypes>
```
as long as flag `-std=c++11` is used.
This should solve issues like https://github.com/facebook/rocksdb/issues/5159
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5402
Differential Revision: D15701195
Pulled By: miasantreble
fbshipit-source-id: 6dac0a05f52aadb55e9728038599d3d2e4b59d03
Summary:
The patch adds a new config option to LRUCacheOptions that enables
users to choose whether to use an adaptive mutex for the LRU block
cache (on platforms where adaptive mutexes are supported). The default
is true if RocksDB is compiled with -DROCKSDB_DEFAULT_TO_ADAPTIVE_MUTEX,
false otherwise.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5054
Differential Revision: D14542749
Pulled By: ltamasi
fbshipit-source-id: 0065715ab6cf91f10444b737fed8c8aee6a8a0d2
Summary:
…ons (#4676)"
This reverts commit b32d087dbb.
`MemoryAllocator` needs to be with `Cache`, since cache entry can
outlive DB and block based table. The cache needs to hold reference to
memory allocator when deleting cache entry.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4697
Differential Revision: D13133490
Pulled By: yiwu-arbug
fbshipit-source-id: 8ef7e8a51263bfd929f892fd062665ff4ce9ce5a
Summary:
Per offline discussion with siying, `MemoryAllocator` and `Cache` should be decouple. The idea is that memory allocator handles memory allocation, while cache handle cache policy.
It is normal that external cache libraries pack couple the two components for better optimization. If we want to integrate with such library in the future, we can make a wrapper of the library implementing both `Cache` and `MemoryAllocator` interface.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4676
Differential Revision: D13047662
Pulled By: yiwu-arbug
fbshipit-source-id: cd42e246d80ab600b4de47d073f7d2db308ce6dd
Summary:
Rename the interface, as it is mean to be a generic interface for memory allocation.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4590
Differential Revision: D10866340
Pulled By: yiwu-arbug
fbshipit-source-id: 85cb753351a40cb856c046aeaa3f3b369eef3d16
Summary:
This is a conceptually simple change, but it touches many files to
pass the allocator through function calls.
We introduce CacheAllocator, which can be used by clients to configure
custom allocator for cache blocks. Our motivation is to hook this up
with folly's `JemallocNodumpAllocator`
(f43ce6d686/folly/experimental/JemallocNodumpAllocator.h),
but there are many other possible use cases.
Additionally, this commit cleans up memory allocation in
`util/compression.h`, making sure that all allocations are wrapped in a
unique_ptr as soon as possible.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4437
Differential Revision: D10132814
Pulled By: yiwu-arbug
fbshipit-source-id: be1343a4b69f6048df127939fea9bbc96969f564
Summary:
This PR supports the group commit of multiple version edit entries corresponding to different column families. Column family drop/creation still cannot be grouped. This PR is a subset of [PR 3752](https://github.com/facebook/rocksdb/pull/3752).
Closes https://github.com/facebook/rocksdb/pull/3944
Differential Revision: D8432536
Pulled By: riversand963
fbshipit-source-id: 8f11bd05193b6c0d9272d82e44b676abfac113cb
Summary:
Instead of __SANITIZE_ADDRESS__ macro, LLVM uses __has_feature(address_sanitzer) to check if ASAN is enabled for the build. I tested it with MySQL sanitizer build that uses RocksDB as a submodule.
Closes https://github.com/facebook/rocksdb/pull/4066
Reviewed By: riversand963
Differential Revision: D8668941
Pulled By: taewookoh
fbshipit-source-id: af4d1da180c1470d257a228f431eebc61490bc36
Summary:
Implement midpoint insertion strategy where new blocks will be insert to the middle of LRU list, then move the head on the first hit in cache.
Closes https://github.com/facebook/rocksdb/pull/3877
Differential Revision: D8100895
Pulled By: yiwu-arbug
fbshipit-source-id: f4bd83cb8be469e5d02072cfc8bd66011391f3da
