Summary:
HyperClockCache is intended to mitigate performance problems under stress conditions (as well as optimizing average-case parallel performance). In LRUCache, the biggest such problem is lock contention when one or a small number of cache entries becomes particularly hot. Regardless of cache sharding, accesses to any particular cache entry are linearized against a single mutex, which is held while each access updates the LRU list. All HCC variants are fully lock/wait-free for accessing blocks already in the cache, which fully mitigates this contention problem.
However, HCC (and CLOCK in general) can exhibit extremely degraded performance under a different stress condition: when no (or almost no) entries in a cache shard are evictable (they are pinned). Unlike LRU which can find any evictable entries immediately (at the cost of more coordination / synchronization on each access), CLOCK has to search for evictable entries. Under the right conditions (almost exclusively MB-scale caches not GB-scale), the CPU cost of each cache miss could fall off a cliff and bog down the whole system.
To effectively mitigate this problem (IMHO), I'm introducing a new default behavior and tuning parameter for HCC, `eviction_effort_cap`. See the comments on the new config parameter in the public API.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12141
Test Plan:
unit test included
## Performance test
We can use cache_bench to validate no regression (CPU and memory) in normal operation, and to measure change in behavior when cache is almost entirely pinned. (TODO: I'm not sure why I had to get the pinned ratio parameter well over 1.0 to see truly bad performance, but the behavior is there.) Build with `make DEBUG_LEVEL=0 USE_CLANG=1 PORTABLE=0 cache_bench`. We also set MALLOC_CONF="narenas:1" for all these runs to essentially remove jemalloc variances from the results, so that the max RSS given by /usr/bin/time is essentially ideal (assuming the allocator minimizes fragmentation and other memory overheads well). Base command reproducing bad behavior:
```
./cache_bench -cache_type=auto_hyper_clock_cache -threads=12 -histograms=0 -pinned_ratio=1.7
```
```
Before, LRU (alternate baseline not exhibiting bad behavior):
Rough parallel ops/sec = 2290997
1088060 maxresident
Before, AutoHCC (bad behavior):
Rough parallel ops/sec = 141011 <- Yes, more than 10x slower
1083932 maxresident
```
Now let us sample a range of values in the solution space:
```
After, AutoHCC, eviction_effort_cap = 1:
Rough parallel ops/sec = 3212586
2402216 maxresident
After, AutoHCC, eviction_effort_cap = 10:
Rough parallel ops/sec = 2371639
1248884 maxresident
After, AutoHCC, eviction_effort_cap = 30:
Rough parallel ops/sec = 1981092
1131596 maxresident
After, AutoHCC, eviction_effort_cap = 100:
Rough parallel ops/sec = 1446188
1090976 maxresident
After, AutoHCC, eviction_effort_cap = 1000:
Rough parallel ops/sec = 549568
1084064 maxresident
```
I looks like `cap=30` is a sweet spot balancing acceptable CPU and memory overheads, so is chosen as the default.
```
Change to -pinned_ratio=0.85
Before, LRU:
Rough parallel ops/sec = 2108373
1078232 maxresident
Before, AutoHCC, averaged over ~20 runs:
Rough parallel ops/sec = 2164910
1077312 maxresident
After, AutoHCC, eviction_effort_cap = 30, averaged over ~20 runs:
Rough parallel ops/sec = 2145542
1077216 maxresident
```
The slight CPU improvement above is consistent with the cap, with no measurable memory overhead under moderate stress.
```
Change to -pinned_ratio=0.25 (low stress)
Before, AutoHCC, averaged over ~20 runs:
Rough parallel ops/sec = 2221149
1076540 maxresident
After, AutoHCC, eviction_effort_cap = 30, averaged over ~20 runs:
Rough parallel ops/sec = 2224521
1076664 maxresident
```
No measurable difference under normal circumstances.
Some tests repeated with FixedHCC, with similar results.
Reviewed By: anand1976
Differential Revision: D52174755
Pulled By: pdillinger
fbshipit-source-id: d278108031b1220c1fa4c89c5a9d34b7cf4ef1b8
Summary:
Example:
```
cache/clock_cache.cc:56:7: error: fallthrough annotation in unreachable code [-Werror,-Wimplicit-fallthrough]
FALLTHROUGH_INTENDED;
^
./port/lang.h:10:30: note: expanded from macro 'FALLTHROUGH_INTENDED'
^
```
In clang < 14, this is annoyingly generated from -Wimplicit-fallthrough, but was changed to -Wunreachable-code-fallthrough (implied by -Wunreachable-code) in clang 14. See https://reviews.llvm.org/D107933 for how this nuisance pattern generated false positives similar to ours in the Linux kernel.
Just to underscore the ridiculousness of this warning, here an error is reported on the annotation, not the call to do_something(), depending on the constexpr value (https://godbolt.org/z/EvxqdPTdr):
```
#include <atomic>
void do_something();
void test(int v) {
switch (v) {
case 1:
if constexpr (std::atomic<long>::is_always_lock_free) {
return;
} else {
do_something();
[[fallthrough]];
}
case 2:
return;
}
}
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12144
Test Plan: Added the warning to our Makefile for USE_CLANG, which reproduced the warning-as-error as shown above, but is now fixed.
Reviewed By: jaykorean
Differential Revision: D52139615
Pulled By: pdillinger
fbshipit-source-id: ba967ae700c0916d1a478bc465cf917633e337d9
Summary:
Sanitize the `secondary_cache` field in the `cache_opts` option of `TieredCacheOptions` to `nullptr` if set by the user. The nvm secondary cache should be directly set in `TieredCacheOptions`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12137
Reviewed By: akankshamahajan15
Differential Revision: D52063817
Pulled By: anand1976
fbshipit-source-id: 255116c665a9b908c8f44109a2d331d4b73e7591
Summary:
There is a bug in the `TieredSecondaryCache` that can result in a false negative. This can happen when a MultiGet does a cache lookup that gets a hit in the `TieredSecondaryCache` local nvm cache tier, and the result is available before MultiGet calls `WaitAll` (i.e the nvm cache `SecondaryCacheResultHandle` `IsReady` returns true).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12134
Test Plan: Add a new unit test in tiered_secondary_cache_test
Reviewed By: akankshamahajan15
Differential Revision: D52023309
Pulled By: anand1976
fbshipit-source-id: e5ae681226a0f12753fecb2f6acc7e5f254ae72b
Summary:
Add some asserts in the `CacheWithSecondaryAdapter` destructor to help debug a crash test failure.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12082
Reviewed By: cbi42
Differential Revision: D51486041
Pulled By: anand1976
fbshipit-source-id: 76537beed31ba27ab9ac8b4ce6deb775629e3be5
Summary:
`CacheWithSecondaryAdapter` can distribute placeholder reservations across the primary and secondary caches. The current implementation of the accounting is quite complicated in order to avoid using a mutex. This may cause the accounting to be slightly off after changes to the cache capacity and ratio, resulting in assertion failures. There's also a bug in the unlikely event that the total reservation exceeds the cache capacity. Furthermore, the current implementation is difficult to reason about.
This PR simplifies it by doing the accounting while holding a mutex. The reservations are processed in 1MB chunks in order to avoid taking a lock too frequently. As a side effect, this also removes the restriction of not allowing to increase the compressed secondary cache capacity after decreasing it to 0.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12059
Test Plan: Existing unit tests, and a new test for capacity increase from 0
Reviewed By: pdillinger
Differential Revision: D51278686
Pulled By: anand1976
fbshipit-source-id: 7e1ad2c50694772997072dd59cab35c93c12ba4f
Summary:
See new atomic.h file comments for motivation.
I have updated HyperClockCache to use the new atomic wrapper, fixing a few cases where an implicit conversion was accidentally used and therefore mixing std::memory_order_seq_cst where release/acquire ordering (or relaxed) was intended. There probably wasn't a real bug because I think all the cases happened to be in single-threaded contexts like constructors/destructors or statistical ops like `GetCapacity()` that don't need any particular ordering constraints.
Recommended follow-up:
* Replace other uses of std::atomic to help keep them safe from bugs.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12051
Test Plan:
Did some local correctness stress testing with cache_bench. Also triggered 15 runs of fbcode_blackbox_crash_test and saw no related failures (just 3 failures in ~CacheWithSecondaryAdapter(), already known)
No performance difference seen before & after running simultaneously:
```
(while ./cache_bench -cache_type=fixed_hyper_clock_cache -populate_cache=0 -cache_size=3000000000 -ops_per_thread=500000 -threads=12 -histograms=0 2>&1 | grep parallel; do :; done) | awk '{ s += $3; c++; print "Avg time: " (s/c);}'
```
... for both fixed_hcc and auto_hcc.
Reviewed By: jowlyzhang
Differential Revision: D51090518
Pulled By: pdillinger
fbshipit-source-id: eeb324facb3185584603f9ea0c4de6f32919a2d7
Summary:
There was some unncessary logic (e.g. a dead assignment to home_shift) left over from earlier revision of the code.
Also, rename confusing ChainRewriteLock::new_head_ / GetNewHead() to saved_head_ / GetSavedHead().
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12052
Test Plan: existing tests
Reviewed By: jowlyzhang
Differential Revision: D51091499
Pulled By: pdillinger
fbshipit-source-id: 4b191b60a2b16085681e59d49c4d97e802869db8
Summary:
This change simplifies some code and logic by introducing a new atomic field that tracks the next slot to grow into. It should offer slightly better performance during the growth phase (not measurable; see Test Plan below) and fix a suspected (but unconfirmed) bug like this:
* Thread 1 is in non-trivial SplitForGrow() with grow_home=n.
* Thread 2 reaches Grow() with grow_home=2n, and waits at the start of SplitForGrow() for the rewrite lock on n. By this point, the head at 2n is marked with the new shift amount but no chain is locked.
* Thread 3 reaches Grow() with grow_home=4n, and waits before SplitForGrow() for the rewrite lock on n. By this point, the head at 4n is marked with the new shift amount but no chain is locked.
* Thread 4 reaches Grow() with grow_home=8n and meets no resistance to proceeding through a SplitForGrow() on an empty chain, permanently missing out on any entries from chain n that should have ended up here.
This is fixed by not updating the shift amount at the grow_home head until we have checked the preconditions that Grow()s feeding into this one have completed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12047
Test Plan:
Some manual cache_bench stress runs, and about 20 triggered runs of fbcode_blackbox_crash_test
No discernible performance difference on this benchmark, running before & after in parallel for a few minutes:
```
(while ./cache_bench -cache_type=auto_hyper_clock_cache -populate_cache=0 -cache_size=3000000000 -ops_per_thread=50000 -threads=12 -histograms=0 2>&1 | grep parallel; do :; done) | awk '{ s += $3; c++; print "Avg time: " (s/c);}'
```
Reviewed By: jowlyzhang
Differential Revision: D51017007
Pulled By: pdillinger
fbshipit-source-id: 5f6d6a6194fc966f94693f3205ed75c87cdad269
Summary:
I have finally tracked down and fixed a bug affecting AutoHCC that was causing CI crash test assertion failures in AutoHCC when using secondary cache, but I was only able to reproduce locally a couple of times, after very long runs/repetitions.
It turns out that the essential feature used by secondary cache to trigger the bug is Insert without keeping a handle, which is otherwise rarely used in RocksDB and not incorporated into cache_bench (also used for targeted correctness stress testing) until this change (new option `-blind_insert_percent`).
The problem was in copying some logic from FixedHCC that makes the entry "sharable" but unreferenced once populated, if no reference is to be saved. The problem in AutoHCC is that we can only add the entry to a chain after it is in the sharable state, and must be removed from the chain while in the "under (de)construction" state and before it is back in the "empty" state. Also, it is possible for Lookup to find entries that are not connected to any chain, by design for efficiency, and for Release to erase_if_last_ref. Therefore, we could have
* Thread 1 starts to Insert a cache entry without keeping ref, and pauses before adding to the chain.
* Thread 2 finds it with Lookup optimizations, and then does Release with `erase_if_last_ref=true` causing it to trigger erasure on the entry. It successfully locks the home chain for the entry and purges any entries pending erasure. It is OK that this entry is not found on the chain, as another thread is allowed to remove it from the chain before we are able to (but after is it marked for (de)construction). And after the purge of the chain, the entry is marked empty.
* Thread 1 resumes in adding the slot (presumed entry) to the home chain for what was being inserted, but that now violates invariants and sets up a race or double-chain-reference as another thread could insert a new entry in the slot and try to insert into a different chain.
This is easily fixed by holding on to a reference until inserted onto the chain.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12046
Test Plan:
As I don't have a reliable local reproducer, I triggered 20 runs of internal CI on fbcode_blackbox_crash_test that were previously failing in AutoHCC with about 1/3 probability, and they all passed.
Also re-enabling AutoHCC in the crash test with this change. (Revert https://github.com/facebook/rocksdb/issues/12000)
Reviewed By: jowlyzhang
Differential Revision: D51016979
Pulled By: pdillinger
fbshipit-source-id: 3840fb829d65b97c779d8aed62a4a4a433aeff2b
Summary:
Mostly things for using cache_bench for stress/correctness testing.
* Make secondary_cache_uri option work with HCC (forgot to update when secondary support was added for HCC)
* Add -pinned_ratio option to keep more than just one entry per thread pinned. This can be important for testing eviction stress.
* Add -vary_capacity_ratio for testing dynamically changing capacity.
Also added some overrides to CacheWrapper to help with diagnostic output.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12041
Test Plan: manual, make check
Reviewed By: jowlyzhang
Differential Revision: D51013430
Pulled By: pdillinger
fbshipit-source-id: 7914adc1218f0afacace05ccd77d3bfb91a878d0
Summary:
In `TieredCache`, the underlying compressed secondary cache is hidden from the user. So we need a way to query the capacity, as well as the portion of cache reservation charged to the compressed secondary cache.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12011
Test Plan: Update the unit tests
Reviewed By: akankshamahajan15
Differential Revision: D50651943
Pulled By: anand1976
fbshipit-source-id: 06d1cb5edb75a790c919bce718e2ff65f5908220
Summary:
and some other small enhancements/fixes:
* The main bug fixed is that in some rare cases, the "published" table size might be smaller than the actual table size. This is a transient state that can happen with concurrent growth that is normally fixed after enough insertions, but if the cache is destroyed soon enough after growth, it could fail to fully destroy some entries and cause assertion failures. We can fix this by detecting the true table size in the destructor.
* Change the "too many iterations" debug threshold from 512 to 768. We might have hit at least one false positive failure. (Failed despite legitimate operation.)
* Added some stronger assertions in some places to aid in debugging.
* Use COERCE_CONTEXT_SWITCH to make behavior of Grow less predictable in terms of thread interleaving. (Might add in more places.) This was useful in reproducing the destructor bug.
* Fix some comments with typos or that were based on earlier revisions of the code.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11988
Test Plan:
Variants of this bug-finding command:
```
USE_CLANG=1 COMPILE_WITH_ASAN=1 COMPILE_WITH_UBSAN=1 COERCE_CONTEXT_SWITCH=1 DEBUG_LEVEL=2 make -j32 cache_bench && while ROCKSDB_DEBUG=1 ./cache_bench -cache_type=auto_hyper_clock_cache -histograms=0 -cache_size=80000000 -threads=32 -populate_cache=0 -ops_per_thread=1000 -num_shard_bits=0; do :; done
```
Reviewed By: jowlyzhang
Differential Revision: D50470318
Pulled By: pdillinger
fbshipit-source-id: d407a8bb0b6d2ddc598a954c319a1640136f12f2
Summary:
Saw this in stress test:
```
db_stress: cache/clock_cache.cc:3152:[...] Assertion `i < 0x2000' failed.
```
The problem is related to Lookups on a chain currently involved in a Grow operation. To avoid Lookup waiting on Grow, Lookup is able to walk a chain whose first part is already migrated and tail is not yet migrated, so is mixed with entries with a different destination home (according to `home_shift`) than what we're looking for. This is fine until we save one of these entries as a safe point in the chain to backtrack to (`read_ref_on_chain`) in case of concurrent modification and end up backtracking to it. In that case, we can get stuck on the wrong destination chain and keep trying to backtrack to an entry that is supposed to be on the correct chain but is not (anymore).
For some reason I haven't quite worked out, I believe it's usually able to recover after some 1000+ looop iterations, so reproducibility depends on the threshold at which we consider a Lookup loop to be too many iterations for a plausibly valid Lookup.
Detecting and working around this case is relatively simple. We can (and must) keep going on the chain but ensure we don't save it as a safe entry to backtrack to.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11948
Test Plan:
The problem could be reproduced in a few minutes with this (debug build):
```
$ while ./cache_bench -cache_type=auto_hyper_clock_cache -histograms=0 -cache_size=80000000 -threads=32 -populate_cache=0 -ops_per_thread=10000 -degenerate_hash_bits=6 -num_shard_bits=0; do :; done
```
At least with a lower threshold on suspiciously high number of iterations. I've lowered the thresholds quite a bit and no longer able to reproduce a failure.
Reviewed By: jowlyzhang
Differential Revision: D50236574
Pulled By: pdillinger
fbshipit-source-id: 2cb54a4e02bb51d5933eea41fcd489ab9d34aa96
Summary:
With the introduction of the `UpdateTieredCache` API, its possible to dynamically change the compressed secondary cache ratio of the total cache capacity. In order to optimize performance, we avoid using a mutex when inserting/releasing placeholder entries, which can result in some inaccuracy in the accounting during the dynamic update. This inaccuracy was causing a runtime error due to an integer underflow in `UpdateCacheReservationRatio`, causing ubsan crash tests to fail. This PR fixes it by explicitly checking for the underflow.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11949
Test Plan:
1. Added a unit test that fails without the fix
2. Run ubsan_crash
Reviewed By: akankshamahajan15
Differential Revision: D50240217
Pulled By: anand1976
fbshipit-source-id: d2f7b79da54eec8b61aec2cc1f2943da5d5847ac
Summary:
Fix the TSAN false positive caused by reading a bool flag without synchronization.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11941
Test Plan: Run tsan crash test locally
Reviewed By: akankshamahajan15
Differential Revision: D50181799
Pulled By: anand1976
fbshipit-source-id: 889e7237e9f3c9452a9df94a0d949db5fe13bb57
Summary:
In https://github.com/facebook/rocksdb/issues/11812, the ```CacheWithSecondaryAdapter::Insert``` calls ```InsertSaved``` on the secondary cache to warm it up with the compressed blocks. This should only be done if its a stacked cache with compressed and nvm cache. If its in-memory compressed only, then don't call ```InsertSaved```.
Tests:
Add a new unit test
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11889
Reviewed By: akankshamahajan15
Differential Revision: D49615758
Pulled By: anand1976
fbshipit-source-id: 156ff968ad014ac319f8840da7a48193e4cebfa9
Summary:
Updating the tiered cache (cache allocated using ```NewTieredCache()```) by calling ```SetCapacity()``` on it was not working properly. The initial creation would set the primary cache capacity to the combined primary and compressed secondary cache capacity. But ```SetCapacity()``` would just set the primary cache capacity, with no way to change the secondary cache capacity. Additionally, the API was confusing, since the primary and compressed secondary capacities would be specified separately during creation, but ```SetCapacity``` took the combined capacity.
With this fix, the user always specifies the total budget and compressed secondary cache ratio on creation. Subsequently, `SetCapacity` will distribute the new capacity across the two caches by the same ratio. The `NewTieredCache` API has been changed to take the total cache capacity (inclusive of both the primary and the compressed secondary cache) and the ratio of total capacity to allocate to the compressed cache. These are specified in `TieredCacheOptions`. Any capacity specified in `LRUCacheOptions`, `HyperClockCacheOptions` and `CompressedSecondaryCacheOptions` is ignored. A new API, `UpdateTieredCache` is provided to dynamically update the total capacity, ratio of compressed cache, and admission policy.
Tests:
New unit tests
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11873
Reviewed By: akankshamahajan15
Differential Revision: D49562250
Pulled By: anand1976
fbshipit-source-id: 57033bc713b68d5da6292207765a6b3dbe539ddf
Summary:
Example crash seen in crash test:
```
db_stress: cache/clock_cache.cc:237: bool rocksdb::clock_cache::{anonymous}::BeginSlotInsert(const rocksdb::clock_cache::ClockHandleBasicData&, rocksdb::clock_cache::ClockHandle&, uint64_t, bool*): Assertion `*already_matches == false' failed.
```
I was intentionally ignoring `already_matches` without resetting it to false for the next call.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11877
Test Plan:
Reproducer no longer reproduces:
```
while ./cache_bench -cache_type=auto_hyper_clock_cache -threads=32 -populate_cache=0 -histograms=0 -report_problems -insert_percent=87 -lookup_insert_percent=2 -skew=10 -ops_per_thread=100 -cache_size=1000000; do echo hi; done
```
Reviewed By: cbi42
Differential Revision: D49562065
Pulled By: pdillinger
fbshipit-source-id: 941062e6eac7a4b56157925b1cf2a0b15ff9cc9d
Summary:
## The Problem
Mark Callaghan found a performance bug in yet-unreleased AutoHCC (which should have been found in my own testing). The observed behavior is very slow insertion performance as the table is growing into a very large structure. The root cause is the precarious combination of linear hashing (indexing into the table while allowing growth) and linear probing (for finding an empty slot to insert into). Naively combined, this is a disaster because in linear hashing, part of the table is twice as dense as first probing location as the rest. Thus, even a modest load factor like 0.6 could cause the dense part of the table to degrade to linear search. The code had a correction for this imbalance, which works in steady-state operation, but failed to account for the concentrating effect of table growth. Specifically, newly-added slots were underpopulated which allowed old slots to become over-populated and degrade to linear search, even in single-threaded operation. Here's an example:
```
./cache_bench -cache_type=auto_hyper_clock_cache -threads=1 -populate_cache=0 -value_bytes=500 -cache_size=3000000000 -histograms=0 -report_problems -ops_per_thread=20000000 -resident_ratio=0.6
```
AutoHCC: Complete in 774.213 s; Rough parallel ops/sec = 25832
FixedHCC: Complete in 19.630 s; Rough parallel ops/sec = 1018840
LRUCache: Complete in 25.842 s; Rough parallel ops/sec = 773947
## The Fix
One small change is apparently sufficient to fix the problem, but I wanted to re-optimize the whole "finding a good empty slot" algorithm to improve safety margins for good performance and to improve typical case performance.
The small change is to track the newly-added slot from Grow in Insert, when applicable, and use that slot for insertion if (a) the home slot is already occupied, and (b) the newly-added slot is empty. This appears to sufficiently load new slots while avoiding over-population of either old or new slots. See `likely_empty_slot`.
However I've also made the logic much more resilient to parts of the table becoming over-populated. I tested a variant that used double hashing instead of linear probing and found that hurt steady-state average-case performance, presumably due to loss of locality in the chains. And even conventional double hashing might not be ideally robust against density skew in the table (still present because of home location bias), because double hashing might choose a small increment that could take a long time to iterate to the under-populated part of the table.
The compromise that seems to bring the best of each approach is this: do linear probing (+1 at a time) within a small bound (chosen bound of 4 based on performance testing) and then fall back on a double-hashing variant if no slot has been found. The double-hashing variant uses a probing increment that is always close to the golden ratio, relative to the table size, so that any under-populated regions of the table can be found relatively quickly, without introducing any additional skew. And the increment is varied slightly to avoid clustering effects that could happen with a fixed increment (regardless of how big it is).
And that leaves us with one remaining problem: the double hashing increment might not be relatively prime to the table size, so the probing sequence might be a cycle that does not cover the full set of slots. To solve this we can use a technique I developed many years ago (probably also developed by others) that simply adds one (in modular arithmetic) whenever we finish a (potentially incomplete) cycle. This is a simple and reasonably efficient way to iterate over all the slots without repetition, regardless of whether the increment is not relatively prime to the table size, or even zero.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11871
Test Plan:
existing correctness tests, especially ClockCacheTest.ClockTableFull
Intended follow-up: make ClockTableFull test more complete for AutoHCC
## Performance
Ignoring old AutoHCC performance, as we established above it could be terrible. FixedHCC and LRUCache are unaffected by this change. All tests below include this change.
### Getting up to size, single thread
(same cache_bench command as above, all three run at same time)
AutoHCC: Complete in 26.724 s; Rough parallel ops/sec = 748400
FixedHCC: Complete in 19.987 s; Rough parallel ops/sec = 1000631
LRUCache: Complete in 28.291 s; Rough parallel ops/sec = 706939
Single-threaded faster than LRUCache (often / sometimes) is good. FixedHCC has an obvious advantage because it starts at full size.
### Multiple threads, steady state, high hit rate ~95%
Using `-threads=10 -populate_cache=1 -ops_per_thread=10000000` and still `-resident_ratio=0.6`
AutoHCC: Complete in 48.778 s; Rough parallel ops/sec = 2050119
FixedHCC: Complete in 46.569 s; Rough parallel ops/sec = 2147329
LRUCache: Complete in 50.537 s; Rough parallel ops/sec = 1978735
### Multiple threads, steady state, low hit rate ~50%
Change to `-resident_ratio=0.2`
AutoHCC: Complete in 49.264 s; Rough parallel ops/sec = 2029884
FixedHCC: Complete in 49.750 s; Rough parallel ops/sec = 2010041
LRUCache: Complete in 53.002 s; Rough parallel ops/sec = 1886713
Don't expect AutoHCC to be consistently faster than FixedHCC, but they are at least similar in these benchmarks.
Reviewed By: jowlyzhang
Differential Revision: D49548534
Pulled By: pdillinger
fbshipit-source-id: 263e4f4d71d0e9a7d91db3795b48fad75408822b
Summary:
This PR implements support for a three tier cache - primary block cache, compressed secondary cache, and a nvm (local flash) secondary cache. This allows more effective utilization of the nvm cache, and minimizes the number of reads from local flash by caching compressed blocks in the compressed secondary cache.
The basic design is as follows -
1. A new secondary cache implementation, ```TieredSecondaryCache```, is introduced. It keeps the compressed and nvm secondary caches and manages the movement of blocks between them and the primary block cache. To setup a three tier cache, we allocate a ```CacheWithSecondaryAdapter```, with a ```TieredSecondaryCache``` instance as the secondary cache.
2. The table reader passes both the uncompressed and compressed block to ```FullTypedCacheInterface::InsertFull```, allowing the block cache to optionally store the compressed block.
3. When there's a miss, the block object is constructed and inserted in the primary cache, and the compressed block is inserted into the nvm cache by calling ```InsertSaved```. This avoids the overhead of recompressing the block, as well as avoiding putting more memory pressure on the compressed secondary cache.
4. When there's a hit in the nvm cache, we attempt to insert the block in the compressed secondary cache and the primary cache, subject to the admission policy of those caches (i.e admit on second access). Blocks/items evicted from any tier are simply discarded.
We can easily implement additional admission policies if desired.
Todo (In a subsequent PR):
1. Add to db_bench and run benchmarks
2. Add to db_stress
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11812
Reviewed By: pdillinger
Differential Revision: D49461842
Pulled By: anand1976
fbshipit-source-id: b40ac1330ef7cd8c12efa0a3ca75128e602e3a0b
Summary:
This PR makes disabling the compressed secondary cache by setting capacity to 0 a bit more efficient. Previously, inserts/lookups would go to the backing LRUCache before getting rejected due to 0 capacity. With this change, insert/lookup would return from ```CompressedSecondaryCache``` itself.
Tests:
Existing tests
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11863
Reviewed By: akankshamahajan15
Differential Revision: D49476248
Pulled By: anand1976
fbshipit-source-id: f0f17a5e3df7d8bfc06709f8f23c1302056ba590
Summary:
This function uses racing reads for heuristic performance improvement. My change in https://github.com/facebook/rocksdb/issues/11792 only worked for clang, not gcc, and gcc does not accurately handle TSAN suppressions. I would have to mark much more code as suppressed than I want to.
So I've taken a different approach: TSAN build does not use the racing reads but substitutes random results, as an extra test that a "correct" value is not needed for correct overall behavior.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11806
Test Plan: manual TSAN builds & tests with cache_bench
Reviewed By: ajkr
Differential Revision: D49100115
Pulled By: pdillinger
fbshipit-source-id: d6d0dfb796d710b953212dd3fc171b6e88fadea1
Summary:
This change add an experimental next-generation HyperClockCache (HCC) with automatic sizing of the underlying hash table. Both the existing version (stable) and the new version (experimental for now) of HCC are available depending on whether an estimated average entry charge is provided in HyperClockCacheOptions.
Internally, we call the two implementations AutoHyperClockCache (new) and FixedHyperClockCache (existing). The performance characteristics and much of the underlying logic are similar enough that AutoHCC is likely to make FixedHCC obsolete, and so it's best considered an evolution of the same technology or solution rather than an alternative. More specifically, both implementations share essentially the same logic for managing the state of individual entries in the cache, including metadata for reference counting and counting clocks for eviction. This metadata, which I like to call the "low-level HCC protocol," includes a read-write lock on entries, but relaxed consistency requirements on the cache (e.g. allowing rare duplication) means high-level cache operations never wait for these low-level per-entry locks. FixedHCC is fully wait-free.
AutoHCC is different in how entries are indexed into an efficient hash table. AutoHCC is "essentially wait-free" as there is no pattern of typical high-level operations on a large cache that can lead to one thread waiting on another to complete some work, though it can happen in some unusual/unlucky cases, or atypical uses such as erasing specific cache keys. Table growth and entry reclamation is more complex in AutoHCC compared to FixedHCC, so uses some localized locking to manage that. AutoHCC uses linear hashing to grow the table as needed, with low latency and to a precise size. AutoHCC depends on anonymous mmap support from the OS (currently verified working on Linux, MacOS, and Windows) to allow the array underlying a hash table to grow in place without wasting resident memory on space reserved but unused. AutoHCC uses a form of chaining while FixedHCC uses open addressing and double hashing.
More specifics:
* In developing this PR, a rare availability bug (minor) was noticed in the existing HCC implementation of Release()+erase_if_last_ref, which is now inherited into AutoHCC. Fixing this without a performance regression will not be simple, so is left for follow-up work.
* Some existing unit tests required adjustment of operational parameters or conditions to work with the new behaviors of AutoHCC. A number of bugs were found and fixed in the validation process, including getting unit tests in good working order.
* Added an option to cache_bench, `-degenerate_hash_bits` for correctness stress testing described below. For this, the tool uses the reverse-engineered hash function for HCC to generate keys in which the specified number of hash bits, in critical positions, have a fixed value. Essentially each degenerate hash bit will half the number of chain heads utilized and double the average chain length.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11738
Test Plan:
unit tests updated, and already added to db crash test. Also
## Correctness
The code includes generous assertions to check for unexpected states, especially at destruction time, so should be able to detect critical concurrency bugs. Less serious "availability bugs" in which cache data is hidden or cleanly lost are more difficult to detect, but also less scary for data correctness (as long as performance is good and the design is sound).
In average operation, the structure is extremely low stress and low contention (see next section) so stressing the corner case logic requires artificially stressing the operating conditions. First, we keep the structure small to increase the number of threads hitting the same chain or entry, and just one cache shard. Second, we artificially degrade the hashing so that chains are much longer than typical, using the new `-degenerate_hash_bits` option to cache_bench. Third, we re-create the structure from scratch frequently in order to exercise the Grow logic repeatedly and to get the benefit of the consistency checks in the structure's destructor in debug builds. For cache_bench this also means disabling the single-threaded "populate cache" step (normally used for steady state performance testing). And of course use many more threads than cores to have many preemptions.
An effective test for working out bugs was this (using debug build of course):
```
while ./cache_bench -cache_type=auto_hyper_clock_cache -histograms=0 -cache_size=8000000 -threads=100 -populate_cache=0 -ops_per_thread=10000 -degenerate_hash_bits=6 -num_shard_bits=0; do :; done
```
Or even smaller cases. This setup has around 27 utilized chains, with around 35 entries each, and yield-waits more than 1 million times per second (very high contention; see next section). I have let this run for hours searching for any lingering issues.
I've also run cache_bench under ASAN, UBSAN, and TSAN.
## Essentially wait free
There is a counter for number of yield() calls when one thread is waiting on another. When we pre-populate the structure in a single thread,
```
./cache_bench -cache_type=auto_hyper_clock_cache -histograms=0 -populate_cache=1 -ops_per_thread=200000 2>&1 | grep Yield
```
We see something on the order of 1 yield call per second across 16 threads, even when we load the system other other jobs (parallel compilation). With -populate_cache=0, there are more yield opportunities with parallel table growth. On an otherwise unloaded system, we still see very small (single digit) yield counts, with a chance of getting into the thousands, and getting into 10s of thousands per second during table growth phase if the system is loaded with other jobs. However, I am not worried about this if performance is still good (see next section).
## Overall performance
Although cache_bench initially suggested performance very close to FixedHCC, there was a very noticeable performance hit under a db_bench setup like used in validating https://github.com/facebook/rocksdb/issues/10626. Much of the difference has been reduced by optimizing Lookup with a "naive" pass that will almost always find entries quickly, and only falling back to the careful Lookup algorithm when not found in the first pass.
Setups (chosen to be sensitive to block cache performance), and compiled with USE_CLANG=1 JEMALLOC=1 PORTABLE=0 DEBUG_LEVEL=0:
```
TEST_TMPDIR=/dev/shm base/db_bench -benchmarks=fillrandom -num=30000000 -disable_wal=1 -bloom_bits=16
```
### No regression on FixedHCC
Running before & after builds at the same time on a 48 core machine.
```
TEST_TMPDIR=/dev/shm /usr/bin/time ./db_bench -benchmarks=readrandom[-X10],block_cache_entry_stats,cache_report_problems -readonly -num=30000000 -bloom_bits=16 -cache_index_and_filter_blocks=1 -cache_size=610000000 -duration 20 -threads=24 -cache_type=fixed_hyper_clock_cache -seed=1234
```
Before:
readrandom [AVG 10 runs] : 847234 (± 8150) ops/sec; 59.2 (± 0.6) MB/sec
703MB max RSS
After:
readrandom [AVG 10 runs] : 851021 (± 7929) ops/sec; 59.5 (± 0.6) MB/sec
706MB max RSS
Probably no material difference.
### Single-threaded performance
Using `[-X2]` and `-threads=1` and `-duration=30`, running all three at the same time:
lru_cache: 55100 ops/sec, then 55862 ops/sec (627MB max RSS)
fixed_hyper_clock_cache: 60496 ops/sec, then 61231 ops/sec (626MB max RSS)
auto_hyper_clock_cache: 47560 ops/sec, then 56081 ops/sec (626MB max RSS)
So AutoHCC has more ramp-up cost in the first pass as the cache grows to the appropriate size. (In single-threaded operation, the parallelizability and per-op low latency of table growth is overall slower.) However, once up to size, its performance is comparable to LRUCache. FixedHCC's lean operations still win overall when a good estimate is available.
If we look at HCC table stats, we can see that this configuration is not favorable to AutoHCC (and I have verified that other memory sizes do not yield substantially different results, until shards are under-sized for the full filters):
FixedHCC:
Slot occupancy stats: Overall 47% (124991/262144), Min/Max/Window = 28%/64%/500, MaxRun{Pos/Neg} = 17/22
AutoHCC:
Slot occupancy stats: Overall 59% (125781/209682), Min/Max/Window = 43%/82%/500, MaxRun{Pos/Neg} = 76/16
Head occupancy stats: Overall 43% (92259/209682), Min/Max/Window = 24%/74%/500, MaxRun{Pos/Neg} = 19/26
Entries at home count: 53350
FixedHCC configuration is relatively good for speed, and not ideal for space utilization. As is typical, AutoHCC has tighter control on metadata usage (209682 x 64 bytes rather than 262144 x 64 bytes), and the higher load factor is slightly worse for speed. LRUCache also has more metadata usage, at 199680 x 96 bytes of tracked metadata (plus roughly another 10% of that untracked in the head pointers), and that metadata is subject to fragmentation.
### Parallel performance, high hit rate
Now using `[-X10]` and `-threads=10`, all three at the same time
lru_cache: [AVG 10 runs] : 263629 (± 1425) ops/sec; 18.4 (± 0.1) MB/sec
655MB max RSS, 97.1% cache hit rate
fixed_hyper_clock_cache: [AVG 10 runs] : 479590 (± 8114) ops/sec; 33.5 (± 0.6) MB/sec
651MB max RSS, 97.1% cache hit rate
auto_hyper_clock_cache: [AVG 10 runs] : 418687 (± 5915) ops/sec; 29.3 (± 0.4) MB/sec
657MB max RSS, 97.1% cache hit rate
Even with just 10-way parallelism for each cache (though 30+/48 cores busy overall), LRUCache is already showing performance degradation, while AutoHCC is in the neighborhood of FixedHCC. And that brings us to the question of how AutoHCC holds up under extreme parallelism, so now independent runs with `-threads=100` (overloading 48 cores).
lru_cache: 438613 ops/sec, 827MB max RSS
fixed_hyper_clock_cache: 1651310 ops/sec, 812MB max RSS
auto_hyper_clock_cache: 1505875 ops/sec, 821MB max RSS (Yield count: 1089 over 30s)
Clearly, AutoHCC holds up extremely well under extreme parallelism, even closing some of the modest performance gap with FixedHCC.
### Parallel performance, low hit rate
To get down to roughly 50% cache hit rate, we use `-cache_index_and_filter_blocks=0 -cache_size=1650000000` with `-threads=10`. Here the extra cost of running counting clock eviction, especially on the chains of AutoHCC, are evident, especially with the lower contention of cache_index_and_filter_blocks=0:
lru_cache: 725231 ops/sec, 1770MB max RSS, 51.3% hit rate
fixed_hyper_clock_cache: 638620 ops/sec, 1765MB max RSS, 50.2% hit rate
auto_hyper_clock_cache: 541018 ops/sec, 1777MB max RSS, 50.8% hit rate
Reviewed By: jowlyzhang
Differential Revision: D48784755
Pulled By: pdillinger
fbshipit-source-id: e79813dc087474ac427637dd282a14fa3011a6e4
Summary:
* Add some options to cache_bench to use JemallocNodumpAllocator
* Make num_shard_bits option use and report cache-specific defaults
* Add a usleep option to sleep between operations, for simulating a workload with more CPU idle/wait time.
* Use const& for JemallocAllocatorOptions, to improve API usability (e.g. can bind to temporary `{}`)
* InstallStackTraceHandler()
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11758
Test Plan: manual
Reviewed By: jowlyzhang
Differential Revision: D48668479
Pulled By: pdillinger
fbshipit-source-id: b6032fbe09444cdb8f1443a5e017d2eea4f6205a
Summary:
This PR implements a new admission policy for the compressed secondary cache, which includes the functionality of the existing policy, and also admits items evicted from the primary block cache with the hit bit set. Effectively, the new policy works as follows -
1. When an item is demoted from the primary cache without a hit, a placeholder is inserted in the compressed cache. A second demotion will insert the full entry.
2. When an item is promoted from the compressed cache to the primary cache for the first time, a placeholder is inserted in the primary. The second promotion inserts the full entry, while erasing it form the compressed cache.
3. If an item is demoted from the primary cache with the hit bit set, it is immediately inserted in the compressed secondary cache.
The ```TieredVolatileCacheOptions``` has been updated with a new option, ```adm_policy```, which allows the policy to be selected.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11713
Reviewed By: pdillinger
Differential Revision: D48444512
Pulled By: anand1976
fbshipit-source-id: b4cbf8c169a88097dff08e36e8bc4b3088de1492
Summary:
* The plan is for AutoHyperClockCache to be selected when HyperClockCacheOptions::estimated_entry_charge == 0, and in that case to use a new configuration option min_avg_entry_charge for determining an extreme case maximum size for the hash table. For the placeholder, a hack is in place in HyperClockCacheOptions::MakeSharedCache() to make the unit tests happy despite the new options not really making sense with the current implementation.
* Mostly updating and refactoring tests to test both the current HCC (internal name FixedHyperClockCache) and a placeholder for the new version (internal name AutoHyperClockCache).
* Simplify some existing tests not to depend directly on cache type.
* Type-parameterize the shard-level unit tests, which unfortunately requires more syntax like `this->` in places for disambiguation.
* Added means of choosing auto_hyper_clock_cache to cache_bench, db_bench, and db_stress, including add to crash test.
* Add another templated class BaseHyperClockCache to reduce future copy-paste
* Added ReportProblems support to cache_bench
* Added a DEBUG-level diagnostic to ReportProblems for the variance in load factor throughout the table, which will become more of a concern with linear hashing to be used in the Auto implementation. Example with current Fixed HCC:
```
2023/08/10-13:41:41.602450 6ac36 [DEBUG] [che/clock_cache.cc:1507] Slot occupancy stats: Overall 49% (129008/262144), Min/Max/Window = 39%/60%/500, MaxRun{Pos/Neg} = 18/17
```
In other words, with overall occupancy of 49%, the lowest across any 500 contiguous cells is 39% and highest 60%. Longest run of occupied is 18 and longest run of unoccupied is 17. This seems consistent with random samples from a uniform distribution.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11692
Test Plan: Shouldn't be any meaningful changes yet to production code or to what is tested, but there is temporary redundancy in testing until the new implementation is plugged in.
Reviewed By: jowlyzhang
Differential Revision: D48247413
Pulled By: pdillinger
fbshipit-source-id: 11541f996d97af403c2e43c92fb67ff22dd0b5da
Summary:
I'm anticipating using the public name HyperClockCache for both the current version with a fixed-size table and the upcoming version with an automatically growing table. However, for simplicity of testing them as substantially distinct implementations, I want to give them distinct internal names, like FixedHyperClockCache and AutoHyperClockCache.
This change anticipates that by renaming to FixedHyperClockCache and assuming for now that all the unit tests run on HCC will run and behave similarly for the automatic HCC. Obviously updates will need to be made, but I'm trying to avoid uninteresting find & replace updates in what will be a large and engineering-heavy PR for AutoHCC
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11676
Test Plan: no behavior change intended, except logging will now use the name FixedHyperClockCache
Reviewed By: ajkr
Differential Revision: D48103165
Pulled By: pdillinger
fbshipit-source-id: a33f1901488fea102164c2318e2f2b156aaba736
Summary:
More code leading up to dynamic HCC.
* Small enhancements to cache_bench
* Extra assertion in Unref
* Improve a CAS loop in ChargeUsageMaybeEvictStrict
* Put load factor constants in appropriate class
* Move `standalone` field to HyperClockTable::HandleImpl because it can be encoded differently in the upcoming dynamic HCC.
* Add a typed version of MemMapping to simplify some future code.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11670
Test Plan: existing tests, unit test added for TypedMemMapping
Reviewed By: jowlyzhang
Differential Revision: D48056464
Pulled By: pdillinger
fbshipit-source-id: 186b7d3105c5d6d2eb6a592369bc10a97ee14a15
Summary:
... used in validating some HyperClockCache development in progress.
* Revamp the "populate cache" step to avoid redundant insertions (very rare in practice) and more consistently approach the desired resident_ratio while maintaining appropriate skew (still not perfect).
* Track and print hit ratio on lookups, to ensure a fair comparison is happening between implementations etc.
* Add an option to disable tracking and printing histograms (lots of output)
* Add an option to specify a random seed (for more reproducibility)
* Remove confusing/redundant "-skewed" option
Uses BitwiseAnd from https://github.com/facebook/rocksdb/issues/11660 (tested there)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11661
Test Plan: manual
Reviewed By: akankshamahajan15, jowlyzhang
Differential Revision: D47937671
Pulled By: pdillinger
fbshipit-source-id: 85a2bb881b1bca4f63e015bac684105fd91c9f35
Summary:
BottomNBits() - there is a single fast instruction for this on x86 since BMI2, but testing with godbolt indicates you need at least GCC 10 for the compiler to choose that instruction from the obvious C++ code. https://godbolt.org/z/5a7Ysd41h
BitwiseAnd() - this is a convenience function that works around the language flaw that the type of the result of x & y is the larger of the two input types, when it should be the smaller. This can save some ugly static_cast.
I expect to use both of these in coming HyperClockCache developments, and have applied them in a couple of places in existing code.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11660
Test Plan: unit tests added
Reviewed By: jowlyzhang
Differential Revision: D47935531
Pulled By: pdillinger
fbshipit-source-id: d148c43a1e51df4a1c549b93aaf2725a3f8d3bd6
Summary:
... ahead of dynamic variant.
* Introduce an Unref function for a common pattern. Cases that were previously using std::memory_order_acq_rel we doing so because we were saving the pre-updated value in case it might be used. Now we are explicitly throwing away the pre-updated value so do not need the acquire semantic, just release.
* Introduce a reusable EvictionData struct and TrackAndReleaseEvictedEntry() function.
* Based on a linter suggesting, use const Func& parameter type instead of Func for templated callable parameters.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11630
Test Plan: existing tests, and performance test with release build of cache_bench. Getting 1-2% difference between before & after from run to run, but inconsistent about which one is faster.
Reviewed By: jowlyzhang
Differential Revision: D47657334
Pulled By: pdillinger
fbshipit-source-id: 5cf2377c0d47a39143b04be6735f98c550e8bdc3
Summary:
An internal user wants to implement a key-aware row cache policy. For that, they need to know the components of the cache key, especially the user key component. With a specialized `RowCache` interface, we will be able to tell them the components so they won't have to make assumptions about our internal key schema.
This PR prepares for the specialized `RowCache` interface by updating the migration plan of https://github.com/facebook/rocksdb/issues/11450. I added a release note for the removed APIs and didn't mention the added ones for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11620
Reviewed By: pdillinger
Differential Revision: D47536962
Pulled By: ajkr
fbshipit-source-id: bbee0fc4ad67fc699a66b8f2b4ea4544dd003691
Summary:
... ahead of planned dynamic HCC variant. This changes
simplifies some logic while still enabling future code sharing between
implementation variants.
Detail: For complicated reasons, using a std::function parameter to
`ConstApplyToEntriesRange` with a lambda argument does not play
nice with templated HandleImpl. An explicit conversion to std::function
would be needed for it to compile. Templating the function type is the
easy work-around.
Also made some functions from https://github.com/facebook/rocksdb/issues/11572 private as recommended
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11609
Test Plan: existing tests
Reviewed By: jowlyzhang
Differential Revision: D47407415
Pulled By: pdillinger
fbshipit-source-id: 0f65954db16335999b78fb7d2563ec627624cef0
Summary:
Stacked on https://github.com/facebook/rocksdb/issues/11572
* Minimize use of std::function and lambdas to minimize chances of
compiler heap-allocating closures (unnecessary stress on allocator). It
appears that converting FindSlot to a template enables inlining the
lambda parameters, avoiding heap allocations.
* Clean up some logic with FindSlot (FIXMEs from https://github.com/facebook/rocksdb/issues/11572)
* Fix handling of rare case of probing all slots, with new unit test.
(Previously Insert would not roll back displacements in that case, which
would kill performance if it were to happen.)
* Add an -early_exit option to cache_bench for gathering memory stats
before deallocation.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11601
Test Plan:
unit test added for probing all slots
## Seeing heap allocations
Run `MALLOC_CONF="stats_print:true" ./cache_bench -cache_type=hyper_clock_cache`
before https://github.com/facebook/rocksdb/issues/11572 vs. after this change. Before, we see this in the
interesting bin statistics:
```
size nrequests
---- ---------
32 578460
64 24340
8192 578460
```
And after:
```
size nrequests
---- ---------
32 (insignificant)
64 24370
8192 579130
```
## Performance test
Build with `make USE_CLANG=1 PORTABLE=0 DEBUG_LEVEL=0 -j32 cache_bench`
Run `./cache_bench -cache_type=hyper_clock_cache -ops_per_thread=5000000`
in before and after configurations, simultaneously:
```
Before: Complete in 33.244 s; Rough parallel ops/sec = 2406442
After: Complete in 32.773 s; Rough parallel ops/sec = 2441019
```
Reviewed By: jowlyzhang
Differential Revision: D47375092
Pulled By: pdillinger
fbshipit-source-id: 46f0f57257ddb374290a0a38c651764ea60ba410
Summary:
Separate out some functionality that will be common to both static and dynamic HCC into BaseClockTable. Table::InsertState and GrowIfNeeded will be used by the dynamic HCC so don't make much sense right now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11572
Test Plan:
existing tests. No functional changes intended.
Performance test in subsequent PR https://github.com/facebook/rocksdb/issues/11601
Reviewed By: jowlyzhang
Differential Revision: D47110496
Pulled By: pdillinger
fbshipit-source-id: 379bd433322a42ea28c0043b41ec24956d21e7aa
Summary:
Infer detected a(n) [Unnecessary Copy Intermediate](https://fbinfer.com/docs/next/all-issue-types#unnecessary copy intermediate) issue. variable &my_secondary_handles is copied unnecessarily into an intermediate on line 268. To avoid the copy, try moving it by calling std::move instead or alternatively change the callee's parameter type to const &.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11566
Reviewed By: akankshamahajan15
Differential Revision: D47057361
Pulled By: ajkr
fbshipit-source-id: bc5d7a71638aecbf976f1a163128b489c9e87fd8
Summary:
This draft PR implements charging of reserved memory, for write buffers, table readers, and other purposes, proportionally to the block cache and the compressed secondary cache. The basic flow of memory reservation is maintained - clients use ```CacheReservationManager``` to request reservations, and ```CacheReservationManager``` inserts placeholder entries, i.e null value and non-zero charge, into the block cache. The ```CacheWithSecondaryAdapter``` wrapper uses its own instance of ```CacheReservationManager``` to keep track of reservations charged to the secondary cache, while the placeholder entries are inserted into the primary block cache. The design is as follows.
When ```CacheWithSecondaryAdapter``` is constructed with the ```distribute_cache_res``` parameter set to true, it manages the entire memory budget across the primary and secondary cache. The secondary cache is assumed to be in memory, such as the ```CompressedSecondaryCache```. When a placeholder entry is inserted by a CacheReservationManager instance to reserve memory, the ```CacheWithSecondaryAdapter```ensures that the reservation is distributed proportionally across the primary/secondary caches.
The primary block cache is initially sized to the sum of the primary cache budget + the secondary cache budget, as follows -
|--------- Primary Cache Configured Capacity -----------|
|---Secondary Cache Budget----|----Primary Cache Budget-----|
A ```ConcurrentCacheReservationManager``` member in the ```CacheWithSecondaryAdapter```, ```pri_cache_res_```, is used to help with tracking the distribution of memory reservations. Initially, it accounts for the entire secondary cache budget as a reservation against the primary cache. This shrinks the usable capacity of the primary cache to the budget that the user originally desired.
|--Reservation for Sec Cache--|-Pri Cache Usable Capacity---|
When a reservation placeholder is inserted into the adapter, it is inserted directly into the primary cache. This means the entire charge of the placeholder is counted against the primary cache. To compensate and count a portion of it against the secondary cache, the secondary cache ```Deflate()``` method is called to shrink it. Since the ```Deflate()``` causes the secondary actual usage to shrink, it is reflected here by releasing an equal amount from the ```pri_cache_res_``` reservation.
For example, if the pri/sec ratio is 50/50, this would be the state after placeholder insertion -
|-Reservation for Sec Cache-|-Pri Cache Usable Capacity-|-R-|
Likewise, when the user inserted placeholder is released, the secondary cache ```Inflate()``` method is called to grow it, and the ```pri_cache_res_``` reservation is increased by an equal amount.
Other alternatives -
1. Another way of implementing this would have been to simply split the user reservation in ```CacheWithSecondaryAdapter``` into primary and secondary components. However, this would require allocating a structure to track the associated secondary cache reservation, which adds some complexity and overhead.
2. Yet another option is to implement the splitting directly in ```CacheReservationManager```. However, there are multiple instances of ```CacheReservationManager``` in a DB instance, making it complicated to keep track of them.
The PR contains the following changes -
1. A new cache allocator, ```NewTieredVolatileCache()```, is defined for allocating a tiered primary block cache and compressed secondary cache. This internally allocates an instance of ```CacheWithSecondaryAdapter```.
3. New interfaces, ```Deflate()``` and ```Inflate()```, are added to the ```SecondaryCache``` interface. The default implementaion returns ```NotSupported``` with overrides in ```CompressedSecondaryCache```.
4. The ```CompressedSecondaryCache``` uses a ```ConcurrentCacheReservationManager``` instance to manage reservations done using ```Inflate()/Deflate()```.
5. The ```CacheWithSecondaryAdapter``` optionally distributes memory reservations across the primary and secondary caches. The primary cache is sized to the total memory budget (primary + secondary), and the capacity allocated to secondary cache is "reserved" against the primary cache. For any subsequent reservations, the primary cache pre-reserved capacity is adjusted.
Benchmarks -
Baseline
```
time ~/rocksdb_anand76/db_bench --db=/dev/shm/comp_cache_res/base --use_existing_db=true --benchmarks="readseq,readwhilewriting" --key_size=32 --value_size=1024 --num=20000000 --threads=32 --bloom_bits=10 --cache_size=30000000000 --use_compressed_secondary_cache=true --compressed_secondary_cache_size=5000000000 --duration=300 --cost_write_buffer_to_cache=true
```
```
readseq : 3.301 micros/op 9694317 ops/sec 66.018 seconds 640000000 operations; 9763.0 MB/s
readwhilewriting : 22.921 micros/op 1396058 ops/sec 300.021 seconds 418846968 operations; 1405.9 MB/s (13068999 of 13068999 found)
real 6m31.052s
user 152m5.660s
sys 26m18.738s
```
With TieredVolatileCache
```
time ~/rocksdb_anand76/db_bench --db=/dev/shm/comp_cache_res/base --use_existing_db=true --benchmarks="readseq,readwhilewriting" --key_size=32 --value_size=1024 --num=20000000 --threads=32 --bloom_bits=10 --cache_size=30000000000 --use_compressed_secondary_cache=true --compressed_secondary_cache_size=5000000000 --duration=300 --cost_write_buffer_to_cache=true --use_tiered_volatile_cache=true
```
```
readseq : 4.064 micros/op 7873915 ops/sec 81.281 seconds 640000000 operations; 7929.7 MB/s
readwhilewriting : 20.944 micros/op 1527827 ops/sec 300.020 seconds 458378968 operations; 1538.6 MB/s (14296999 of 14296999 found)
real 6m42.743s
user 157m58.972s
sys 33m16.671
```
```
readseq : 3.484 micros/op 9184967 ops/sec 69.679 seconds 640000000 operations; 9250.0 MB/s
readwhilewriting : 21.261 micros/op 1505035 ops/sec 300.024 seconds 451545968 operations; 1515.7 MB/s (14101999 of 14101999 found)
real 6m31.469s
user 155m16.570s
sys 27m47.834s
```
ToDo -
1. Add to db_stress
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11449
Reviewed By: pdillinger
Differential Revision: D46197388
Pulled By: anand1976
fbshipit-source-id: 42d16f0254df683db4929db20d06ff26030e90df
Summary:
Add two type aliases for Cache: BlockCache and GeneralCache, and add LRUCacheOptions::MakeSharedGeneralCache(). This will ease upgrade to an intended future change to separate the cache API between block cache and other (general) uses, including row cache. Separating the APIs will make it easier to expose more details of block caching for customization. For example, it would be nice to pass the file unique ID and offset as the logical cache key instead of using a Slice, which could facilitate some file-specific customizations in block cache. This would also make it clear that HyperClockCache is not usable as a general cache, because it can only deal with fixed-size block cache keys.
block_cache, row_cache, and blob_cache are the uses of Cache in the public API. blob_cache should be able to use BlockCache while row_cache is a GeneralCache user, as its keys are of arbitrary size.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11450
Test Plan: see updated unit test.
Reviewed By: anand1976
Differential Revision: D45882067
Pulled By: pdillinger
fbshipit-source-id: ff5d9f0b644f87ae337a29a7728ce3ed07b2a4b2
Summary:
In IDE navigation I find it annoying that there are two statistics.h files (etc.) and often land on the wrong one. Here I migrate several headers to use the blah.h <- blah_impl.h <- blah.cc idiom. Although clang-format wants "blah.h" to be the top include for "blah.cc", I think overall this is an improvement.
No public API changes.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11408
Test Plan: existing tests
Reviewed By: ltamasi
Differential Revision: D45456696
Pulled By: pdillinger
fbshipit-source-id: 809d931253f3272c908cf5facf7e1d32fc507373
Summary:
See motivation and description in new ShardedCacheOptions::hash_seed option.
Updated db_bench so that its seed param is used for the cache hash seed.
Made its code more safe to ensure seed is set before use.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11391
Test Plan:
unit tests added / updated
**Performance** - no discernible difference seen running cache_bench repeatedly before & after. With lru_cache and hyper_clock_cache.
Reviewed By: hx235
Differential Revision: D45557797
Pulled By: pdillinger
fbshipit-source-id: 40bf4da6d66f9d41a8a0eb8e5cf4246a4aa07934
Summary:
For better clarity, encouraging more options explicitly specified using fields rather than positionally via constructor parameter lists. Simplifies code maintenance as new fields are added. Deprecate some cases of the confusing pattern of NewWhatever() functions returning shared_ptr.
Net reduction of about 70 source code lines (including comments).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11386
Test Plan: existing tests
Reviewed By: ajkr
Differential Revision: D45059075
Pulled By: pdillinger
fbshipit-source-id: d53fa09b268024f9c55254bb973b6c69feebf41a
Summary:
I previously misread or misinterpreted API contracts for SecondaryCache and this should correct the record. (Follow-up item from https://github.com/facebook/rocksdb/issues/11301)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11316
Test Plan: comments only
Reviewed By: anand1976
Differential Revision: D44245107
Pulled By: pdillinger
fbshipit-source-id: 3f8ddec150674b75728f1730f99b963bbf7b76e7
Summary:
Internally refactors SecondaryCache integration out of LRUCache specifically and into a wrapper/adapter class that works with various Cache implementations. Notably, this relies on separating the notion of async lookup handles from other cache handles, so that HyperClockCache doesn't have to deal with the problem of allocating handles from the hash table for lookups that might fail anyway, and might be on the same key without support for coalescing. (LRUCache's hash table can incorporate previously allocated handles thanks to its pointer indirection.) Specifically, I'm worried about the case in which hundreds of threads try to access the same block and probing in the hash table degrades to linear search on the pile of entries with the same key.
This change is a big step in the direction of supporting stacked SecondaryCaches, but there are obstacles to completing that. Especially, there is no SecondaryCache hook for evictions to pass from one to the next. It has been proposed that evictions be transmitted simply as the persisted data (as in SaveToCallback), but given the current structure provided by the CacheItemHelpers, that would require an extra copy of the block data, because there's intentionally no way to ask for a contiguous Slice of the data (to allow for flexibility in storage). `AsyncLookupHandle` and the re-worked `WaitAll()` should be essentially prepared for stacked SecondaryCaches, but several "TODO with stacked secondaries" issues remain in various places.
It could be argued that the stacking instead be done as a SecondaryCache adapter that wraps two (or more) SecondaryCaches, but at least with the current API that would require an extra heap allocation on SecondaryCache Lookup for a wrapper SecondaryCacheResultHandle that can transfer a Lookup between secondaries. We could also consider trying to unify the Cache and SecondaryCache APIs, though that might be difficult if `AsyncLookupHandle` is kept a fixed struct.
## cache.h (public API)
Moves `secondary_cache` option from LRUCacheOptions to ShardedCacheOptions so that it is applicable to HyperClockCache.
## advanced_cache.h (advanced public API)
* Add `Cache::CreateStandalone()` so that the SecondaryCache support wrapper can use it.
* Add `SetEvictionCallback()` / `eviction_callback_` so that the SecondaryCache support wrapper can use it. Only a single callback is supported for efficiency. If there is ever a need for more than one, hopefully that can be handled with a broadcast callback wrapper.
These are essentially the two "extra" pieces of `Cache` for pulling out specific SecondaryCache support from the `Cache` implementation. I think it's a good trade-off as these are reasonable, limited, and reusable "cut points" into the `Cache` implementations.
* Remove async capability from standard `Lookup()` (getting rid of awkward restrictions on pending Handles) and add `AsyncLookupHandle` and `StartAsyncLookup()`. As noted in the comments, the full struct of `AsyncLookupHandle` is exposed so that it can be stack allocated, for efficiency, though more data is being copied around than before, which could impact performance. (Lookup info -> AsyncLookupHandle -> Handle vs. Lookup info -> Handle)
I could foresee a future in which a Cache internally saves a pointer to the AsyncLookupHandle, which means it's dangerous to allow it to be copyable or even movable. It also means it's not compatible with std::vector (which I don't like requiring as an API parameter anyway), so `WaitAll()` expects any contiguous array of AsyncLookupHandles. I believe this is best for common case efficiency, while behaving well in other cases also. For example, `WaitAll()` has no effect on default-constructed AsyncLookupHandles, which look like a completed cache miss.
## cacheable_entry.h
A couple of functions are obsolete because Cache::Handle can no longer be pending.
## cache.cc
Provides default implementations for new or revamped Cache functions, especially appropriate for non-blocking caches.
## secondary_cache_adapter.{h,cc}
The full details of the Cache wrapper adding SecondaryCache support. Essentially replicates the SecondaryCache handling that was in LRUCache, but obviously refactored. There is a bit of logic duplication, where Lookup() is essentially a manually optimized version of StartAsyncLookup() and Wait(), but it's roughly a dozen lines of code.
## sharded_cache.h, typed_cache.h, charged_cache.{h,cc}, sim_cache.cc
Simply updated for Cache API changes.
## lru_cache.{h,cc}
Carefully remove SecondaryCache logic, implement `CreateStandalone` and eviction handler functionality.
## clock_cache.{h,cc}
Expose existing `CreateStandalone` functionality, add eviction handler functionality. Light refactoring.
## block_based_table_reader*
Mostly re-worked the only usage of async Lookup, which is in BlockBasedTable::MultiGet. Used arrays in place of autovector in some places for efficiency. Simplified some logic by not trying to process some cache results before they're all ready.
Created new function `BlockBasedTable::GetCachePriority()` to reduce some pre-existing code duplication (and avoid making it worse).
Fixed at least one small bug from the prior confusing mixture of async and sync Lookups. In MaybeReadBlockAndLoadToCache(), called by RetrieveBlock(), called by MultiGet() with wait=false, is_cache_hit for the block_cache_tracer entry would not be set to true if the handle was pending after Lookup and before Wait.
## Intended follow-up work
* Figure out if there are any missing stats or block_cache_tracer work in refactored BlockBasedTable::MultiGet
* Stacked secondary caches (see above discussion)
* See if we can make up for the small MultiGet performance regression.
* Study more performance with SecondaryCache
* Items evicted from over-full LRUCache in Release were not being demoted to SecondaryCache, and still aren't to minimize unit test churn. Ideally they would be demoted, but it's an exceptional case so not a big deal.
* Use CreateStandalone for cache reservations (save unnecessary hash table operations). Not a big deal, but worthy cleanup.
* Somehow I got the contract for SecondaryCache::Insert wrong in #10945. (Doesn't take ownership!) That API comment needs to be fixed, but didn't want to mingle that in here.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11301
Test Plan:
## Unit tests
Generally updated to include HCC in SecondaryCache tests, though HyperClockCache has some different, less strict behaviors that leads to some tests not really being set up to work with it. Some of the tests remain disabled with it, but I think we have good coverage without them.
## Crash/stress test
Updated to use the new combination.
## Performance
First, let's check for regression on caches without secondary cache configured. Adding support for the eviction callback is likely to have a tiny effect, but it shouldn't be worrisome. LRUCache could benefit slightly from less logic around SecondaryCache handling. We can test with cache_bench default settings, built with DEBUG_LEVEL=0 and PORTABLE=0.
```
(while :; do base/cache_bench --cache_type=hyper_clock_cache | grep Rough; done) | awk '{ sum += $9; count++; print $0; print "Average: " int(sum / count) }'
```
**Before** this and #11299 (which could also have a small effect), running for about an hour, before & after running concurrently for each cache type:
HyperClockCache: 3168662 (average parallel ops/sec)
LRUCache: 2940127
**After** this and #11299, running for about an hour:
HyperClockCache: 3164862 (average parallel ops/sec) (0.12% slower)
LRUCache: 2940928 (0.03% faster)
This is an acceptable difference IMHO.
Next, let's consider essentially the worst case of new CPU overhead affecting overall performance. MultiGet uses the async lookup interface regardless of whether SecondaryCache or folly are used. We can configure a benchmark where all block cache queries are for data blocks, and all are hits.
Create DB and test (before and after tests running simultaneously):
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=30000000 -disable_wal=1 -bloom_bits=16
TEST_TMPDIR=/dev/shm base/db_bench -benchmarks=multireadrandom[-X30] -readonly -multiread_batched -batch_size=32 -num=30000000 -bloom_bits=16 -cache_size=6789000000 -duration 20 -threads=16
```
**Before**:
multireadrandom [AVG 30 runs] : 3444202 (± 57049) ops/sec; 240.9 (± 4.0) MB/sec
multireadrandom [MEDIAN 30 runs] : 3514443 ops/sec; 245.8 MB/sec
**After**:
multireadrandom [AVG 30 runs] : 3291022 (± 58851) ops/sec; 230.2 (± 4.1) MB/sec
multireadrandom [MEDIAN 30 runs] : 3366179 ops/sec; 235.4 MB/sec
So that's roughly a 3% regression, on kind of a *worst case* test of MultiGet CPU. Similar story with HyperClockCache:
**Before**:
multireadrandom [AVG 30 runs] : 3933777 (± 41840) ops/sec; 275.1 (± 2.9) MB/sec
multireadrandom [MEDIAN 30 runs] : 3970667 ops/sec; 277.7 MB/sec
**After**:
multireadrandom [AVG 30 runs] : 3755338 (± 30391) ops/sec; 262.6 (± 2.1) MB/sec
multireadrandom [MEDIAN 30 runs] : 3785696 ops/sec; 264.8 MB/sec
Roughly a 4-5% regression. Not ideal, but not the whole story, fortunately.
Let's also look at Get() in db_bench:
```
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=readrandom[-X30] -readonly -num=30000000 -bloom_bits=16 -cache_size=6789000000 -duration 20 -threads=16
```
**Before**:
readrandom [AVG 30 runs] : 2198685 (± 13412) ops/sec; 153.8 (± 0.9) MB/sec
readrandom [MEDIAN 30 runs] : 2209498 ops/sec; 154.5 MB/sec
**After**:
readrandom [AVG 30 runs] : 2292814 (± 43508) ops/sec; 160.3 (± 3.0) MB/sec
readrandom [MEDIAN 30 runs] : 2365181 ops/sec; 165.4 MB/sec
That's showing roughly a 4% improvement, perhaps because of the secondary cache code that is no longer part of LRUCache. But weirdly, HyperClockCache is also showing 2-3% improvement:
**Before**:
readrandom [AVG 30 runs] : 2272333 (± 9992) ops/sec; 158.9 (± 0.7) MB/sec
readrandom [MEDIAN 30 runs] : 2273239 ops/sec; 159.0 MB/sec
**After**:
readrandom [AVG 30 runs] : 2332407 (± 11252) ops/sec; 163.1 (± 0.8) MB/sec
readrandom [MEDIAN 30 runs] : 2335329 ops/sec; 163.3 MB/sec
Reviewed By: ltamasi
Differential Revision: D44177044
Pulled By: pdillinger
fbshipit-source-id: e808e48ff3fe2f792a79841ba617be98e48689f5
Summary:
In preparation for factoring secondary cache support out of individual Cache implementations, we can get rid of the "in secondary cache" flag on entries through a workable hack: when an entry is promoted from secondary, it is inserted in primary using a helper that lacks secondary cache support, thus preventing re-insertion into secondary cache through existing logic.
This adds to the complexity of building CacheItemHelpers, because you always have to be able to get to an equivalent helper without secondary cache support, but that complexity is reasonably isolated within RocksDB typed_cache.h and test code.
gcc-7 seems to have problems with constexpr constructor referencing `this` so removed constexpr support on CacheItemHelper.
Also refactored some related test code to share common code / functionality.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11299
Test Plan: existing tests
Reviewed By: anand1976
Differential Revision: D44101453
Pulled By: pdillinger
fbshipit-source-id: 7a59d0a3938ee40159c90c3e65d7004f6a272345
