cuckoocache_tests.cpp raw

   1  // Copyright (c) 2012-present The Bitcoin Core developers
   2  // Distributed under the MIT software license, see the accompanying
   3  // file COPYING or http://www.opensource.org/licenses/mit-license.php.
   4  
   5  #include <cuckoocache.h>
   6  #include <random.h>
   7  #include <script/sigcache.h>
   8  #include <test/util/random.h>
   9  #include <test/util/setup_common.h>
  10  #include <util/byte_units.h>
  11  
  12  #include <boost/test/unit_test.hpp>
  13  
  14  #include <deque>
  15  #include <mutex>
  16  #include <shared_mutex>
  17  #include <thread>
  18  #include <vector>
  19  
  20  /** Test Suite for CuckooCache
  21   *
  22   *  1. All tests should have a deterministic result (using insecure rand
  23   *  with deterministic seeds)
  24   *  2. Some test methods are templated to allow for easier testing
  25   *  against new versions / comparing
  26   *  3. Results should be treated as a regression test, i.e., did the behavior
  27   *  change significantly from what was expected. This can be OK, depending on
  28   *  the nature of the change, but requires updating the tests to reflect the new
  29   *  expected behavior. For example improving the hit rate may cause some tests
  30   *  using BOOST_CHECK_CLOSE to fail.
  31   *
  32   */
  33  BOOST_FIXTURE_TEST_SUITE(cuckoocache_tests, BasicTestingSetup);
  34  
  35  /* Test that no values not inserted into the cache are read out of it.
  36   *
  37   * There are no repeats in the first 200000 m_rng.rand256() calls
  38   */
  39  BOOST_AUTO_TEST_CASE(test_cuckoocache_no_fakes)
  40  {
  41      SeedRandomForTest(SeedRand::ZEROS);
  42      CuckooCache::cache<uint256, SignatureCacheHasher> cc{};
  43      cc.setup_bytes(4_MiB);
  44      for (int x = 0; x < 100000; ++x) {
  45          cc.insert(m_rng.rand256());
  46      }
  47      for (int x = 0; x < 100000; ++x) {
  48          BOOST_CHECK(!cc.contains(m_rng.rand256(), false));
  49      }
  50  };
  51  
  52  struct HitRateTest : BasicTestingSetup {
  53  /** This helper returns the hit rate when bytes*load worth of entries are
  54   * inserted into a bytes sized cache
  55   */
  56  template <typename Cache>
  57  double test_cache(size_t bytes, double load)
  58  {
  59      SeedRandomForTest(SeedRand::ZEROS);
  60      std::vector<uint256> hashes;
  61      Cache set{};
  62      set.setup_bytes(bytes);
  63      uint32_t n_insert = static_cast<uint32_t>(load * (bytes / sizeof(uint256)));
  64      hashes.resize(n_insert);
  65      for (uint32_t i = 0; i < n_insert; ++i) {
  66          uint32_t* ptr = (uint32_t*)hashes[i].begin();
  67          for (uint8_t j = 0; j < 8; ++j)
  68              *(ptr++) = m_rng.rand32();
  69      }
  70      /** We make a copy of the hashes because future optimizations of the
  71       * cuckoocache may overwrite the inserted element, so the test is
  72       * "future proofed".
  73       */
  74      std::vector<uint256> hashes_insert_copy = hashes;
  75      /** Do the insert */
  76      for (const uint256& h : hashes_insert_copy)
  77          set.insert(h);
  78      /** Count the hits */
  79      uint32_t count = 0;
  80      for (const uint256& h : hashes)
  81          count += set.contains(h, false);
  82      double hit_rate = ((double)count) / ((double)n_insert);
  83      return hit_rate;
  84  }
  85  
  86  /** The normalized hit rate for a given load.
  87   *
  88   * The semantics are a little confusing, so please see the below
  89   * explanation.
  90   *
  91   * Examples:
  92   *
  93   * 1. at load 0.5, we expect a perfect hit rate, so we multiply by
  94   * 1.0
  95   * 2. at load 2.0, we expect to see half the entries, so a perfect hit rate
  96   * would be 0.5. Therefore, if we see a hit rate of 0.4, 0.4*2.0 = 0.8 is the
  97   * normalized hit rate.
  98   *
  99   * This is basically the right semantics, but has a bit of a glitch depending on
 100   * how you measure around load 1.0 as after load 1.0 your normalized hit rate
 101   * becomes effectively perfect, ignoring freshness.
 102   */
 103  static double normalize_hit_rate(double hits, double load)
 104  {
 105      return hits * std::max(load, 1.0);
 106  }
 107  }; // struct HitRateTest
 108  
 109  /** Check the hit rate on loads ranging from 0.1 to 1.6 */
 110  BOOST_FIXTURE_TEST_CASE(cuckoocache_hit_rate_ok, HitRateTest)
 111  {
 112      /** Arbitrarily selected Hit Rate threshold that happens to work for this test
 113       * as a lower bound on performance.
 114       */
 115      double HitRateThresh = 0.98;
 116      for (double load = 0.1; load < 2; load *= 2) {
 117          double hits = test_cache<CuckooCache::cache<uint256, SignatureCacheHasher>>(4_MiB, load);
 118          BOOST_CHECK(normalize_hit_rate(hits, load) > HitRateThresh);
 119      }
 120  }
 121  
 122  
 123  struct EraseTest : BasicTestingSetup {
 124  /** This helper checks that erased elements are preferentially inserted onto and
 125   * that the hit rate of "fresher" keys is reasonable*/
 126  template <typename Cache>
 127  void test_cache_erase(size_t bytes)
 128  {
 129      double load = 1;
 130      SeedRandomForTest(SeedRand::ZEROS);
 131      std::vector<uint256> hashes;
 132      Cache set{};
 133      set.setup_bytes(bytes);
 134      uint32_t n_insert = static_cast<uint32_t>(load * (bytes / sizeof(uint256)));
 135      hashes.resize(n_insert);
 136      for (uint32_t i = 0; i < n_insert; ++i) {
 137          uint32_t* ptr = (uint32_t*)hashes[i].begin();
 138          for (uint8_t j = 0; j < 8; ++j)
 139              *(ptr++) = m_rng.rand32();
 140      }
 141      /** We make a copy of the hashes because future optimizations of the
 142       * cuckoocache may overwrite the inserted element, so the test is
 143       * "future proofed".
 144       */
 145      std::vector<uint256> hashes_insert_copy = hashes;
 146  
 147      /** Insert the first half */
 148      for (uint32_t i = 0; i < (n_insert / 2); ++i)
 149          set.insert(hashes_insert_copy[i]);
 150      /** Erase the first quarter */
 151      for (uint32_t i = 0; i < (n_insert / 4); ++i)
 152          BOOST_CHECK(set.contains(hashes[i], true));
 153      /** Insert the second half */
 154      for (uint32_t i = (n_insert / 2); i < n_insert; ++i)
 155          set.insert(hashes_insert_copy[i]);
 156  
 157      /** elements that we marked as erased but are still there */
 158      size_t count_erased_but_contained = 0;
 159      /** elements that we did not erase but are older */
 160      size_t count_stale = 0;
 161      /** elements that were most recently inserted */
 162      size_t count_fresh = 0;
 163  
 164      for (uint32_t i = 0; i < (n_insert / 4); ++i)
 165          count_erased_but_contained += set.contains(hashes[i], false);
 166      for (uint32_t i = (n_insert / 4); i < (n_insert / 2); ++i)
 167          count_stale += set.contains(hashes[i], false);
 168      for (uint32_t i = (n_insert / 2); i < n_insert; ++i)
 169          count_fresh += set.contains(hashes[i], false);
 170  
 171      double hit_rate_erased_but_contained = double(count_erased_but_contained) / (double(n_insert) / 4.0);
 172      double hit_rate_stale = double(count_stale) / (double(n_insert) / 4.0);
 173      double hit_rate_fresh = double(count_fresh) / (double(n_insert) / 2.0);
 174  
 175      // Check that our hit_rate_fresh is perfect
 176      BOOST_CHECK_EQUAL(hit_rate_fresh, 1.0);
 177      // Check that we have a more than 2x better hit rate on stale elements than
 178      // erased elements.
 179      BOOST_CHECK(hit_rate_stale > 2 * hit_rate_erased_but_contained);
 180  }
 181  }; // struct EraseTest
 182  
 183  BOOST_FIXTURE_TEST_CASE(cuckoocache_erase_ok, EraseTest)
 184  {
 185      test_cache_erase<CuckooCache::cache<uint256, SignatureCacheHasher>>(4_MiB);
 186  }
 187  
 188  struct EraseParallelTest : BasicTestingSetup {
 189  template <typename Cache>
 190  void test_cache_erase_parallel(size_t bytes)
 191  {
 192      double load = 1;
 193      SeedRandomForTest(SeedRand::ZEROS);
 194      std::vector<uint256> hashes;
 195      Cache set{};
 196      set.setup_bytes(bytes);
 197      uint32_t n_insert = static_cast<uint32_t>(load * (bytes / sizeof(uint256)));
 198      hashes.resize(n_insert);
 199      for (uint32_t i = 0; i < n_insert; ++i) {
 200          uint32_t* ptr = (uint32_t*)hashes[i].begin();
 201          for (uint8_t j = 0; j < 8; ++j)
 202              *(ptr++) = m_rng.rand32();
 203      }
 204      /** We make a copy of the hashes because future optimizations of the
 205       * cuckoocache may overwrite the inserted element, so the test is
 206       * "future proofed".
 207       */
 208      std::vector<uint256> hashes_insert_copy = hashes;
 209      std::shared_mutex mtx;
 210  
 211      {
 212          /** Grab lock to make sure we release inserts */
 213          std::unique_lock<std::shared_mutex> l(mtx);
 214          /** Insert the first half */
 215          for (uint32_t i = 0; i < (n_insert / 2); ++i)
 216              set.insert(hashes_insert_copy[i]);
 217      }
 218  
 219      /** Spin up 3 threads to run contains with erase.
 220       */
 221      std::vector<std::thread> threads;
 222      threads.reserve(3);
 223      /** Erase the first quarter */
 224      for (uint32_t x = 0; x < 3; ++x)
 225          /** Each thread is emplaced with x copy-by-value
 226          */
 227          threads.emplace_back([&, x] {
 228              std::shared_lock<std::shared_mutex> l(mtx);
 229              size_t ntodo = (n_insert/4)/3;
 230              size_t start = ntodo*x;
 231              size_t end = ntodo*(x+1);
 232              for (uint32_t i = start; i < end; ++i) {
 233                  bool contains = set.contains(hashes[i], true);
 234                  assert(contains);
 235              }
 236          });
 237  
 238      /** Wait for all threads to finish
 239       */
 240      for (std::thread& t : threads)
 241          t.join();
 242      /** Grab lock to make sure we observe erases */
 243      std::unique_lock<std::shared_mutex> l(mtx);
 244      /** Insert the second half */
 245      for (uint32_t i = (n_insert / 2); i < n_insert; ++i)
 246          set.insert(hashes_insert_copy[i]);
 247  
 248      /** elements that we marked erased but that are still there */
 249      size_t count_erased_but_contained = 0;
 250      /** elements that we did not erase but are older */
 251      size_t count_stale = 0;
 252      /** elements that were most recently inserted */
 253      size_t count_fresh = 0;
 254  
 255      for (uint32_t i = 0; i < (n_insert / 4); ++i)
 256          count_erased_but_contained += set.contains(hashes[i], false);
 257      for (uint32_t i = (n_insert / 4); i < (n_insert / 2); ++i)
 258          count_stale += set.contains(hashes[i], false);
 259      for (uint32_t i = (n_insert / 2); i < n_insert; ++i)
 260          count_fresh += set.contains(hashes[i], false);
 261  
 262      double hit_rate_erased_but_contained = double(count_erased_but_contained) / (double(n_insert) / 4.0);
 263      double hit_rate_stale = double(count_stale) / (double(n_insert) / 4.0);
 264      double hit_rate_fresh = double(count_fresh) / (double(n_insert) / 2.0);
 265  
 266      // Check that our hit_rate_fresh is perfect
 267      BOOST_CHECK_EQUAL(hit_rate_fresh, 1.0);
 268      // Check that we have a more than 2x better hit rate on stale elements than
 269      // erased elements.
 270      BOOST_CHECK(hit_rate_stale > 2 * hit_rate_erased_but_contained);
 271  }
 272  }; // struct EraseParallelTest
 273  BOOST_FIXTURE_TEST_CASE(cuckoocache_erase_parallel_ok, EraseParallelTest)
 274  {
 275      test_cache_erase_parallel<CuckooCache::cache<uint256, SignatureCacheHasher>>(4_MiB);
 276  }
 277  
 278  
 279  struct GenerationsTest : BasicTestingSetup {
 280  template <typename Cache>
 281  void test_cache_generations()
 282  {
 283      // This test checks that for a simulation of network activity, the fresh hit
 284      // rate is never below 99%, and the number of times that it is worse than
 285      // 99.9% are less than 1% of the time.
 286      double min_hit_rate = 0.99;
 287      double tight_hit_rate = 0.999;
 288      double max_rate_less_than_tight_hit_rate = 0.01;
 289      // A cache that meets this specification is therefore shown to have a hit
 290      // rate of at least tight_hit_rate * (1 - max_rate_less_than_tight_hit_rate) +
 291      // min_hit_rate*max_rate_less_than_tight_hit_rate = 0.999*99%+0.99*1% == 99.89%
 292      // hit rate with low variance.
 293  
 294      // We use deterministic values, but this test has also passed on many
 295      // iterations with non-deterministic values, so it isn't "overfit" to the
 296      // specific entropy in FastRandomContext(true) and implementation of the
 297      // cache.
 298      SeedRandomForTest(SeedRand::ZEROS);
 299  
 300      // block_activity models a chunk of network activity. n_insert elements are
 301      // added to the cache. The first and last n/4 are stored for removal later
 302      // and the middle n/2 are not stored. This models a network which uses half
 303      // the signatures of recently (since the last block) added transactions
 304      // immediately and never uses the other half.
 305      struct block_activity {
 306          std::vector<uint256> reads;
 307          block_activity(uint32_t n_insert, FastRandomContext& rng, Cache& c)
 308          {
 309              std::vector<uint256> inserts;
 310              inserts.resize(n_insert);
 311              reads.reserve(n_insert / 2);
 312              for (uint32_t i = 0; i < n_insert; ++i) {
 313                  uint32_t* ptr = (uint32_t*)inserts[i].begin();
 314                  for (uint8_t j = 0; j < 8; ++j)
 315                      *(ptr++) = rng.rand32();
 316              }
 317              for (uint32_t i = 0; i < n_insert / 4; ++i)
 318                  reads.push_back(inserts[i]);
 319              for (uint32_t i = n_insert - (n_insert / 4); i < n_insert; ++i)
 320                  reads.push_back(inserts[i]);
 321              for (const auto& h : inserts)
 322                  c.insert(h);
 323          }
 324      };
 325  
 326      const uint32_t BLOCK_SIZE = 1000;
 327      // We expect window size 60 to perform reasonably given that each epoch
 328      // stores 45% of the cache size (~472k).
 329      const uint32_t WINDOW_SIZE = 60;
 330      const uint32_t POP_AMOUNT = (BLOCK_SIZE / WINDOW_SIZE) / 2;
 331      const double load = 10;
 332      const size_t bytes{4_MiB};
 333      const uint32_t n_insert = static_cast<uint32_t>(load * (bytes / sizeof(uint256)));
 334  
 335      std::vector<block_activity> hashes;
 336      Cache set{};
 337      set.setup_bytes(bytes);
 338      hashes.reserve(n_insert / BLOCK_SIZE);
 339      std::deque<block_activity> last_few;
 340      uint32_t out_of_tight_tolerance = 0;
 341      uint32_t total = n_insert / BLOCK_SIZE;
 342      // we use the deque last_few to model a sliding window of blocks. at each
 343      // step, each of the last WINDOW_SIZE block_activities checks the cache for
 344      // POP_AMOUNT of the hashes that they inserted, and marks these erased.
 345      for (uint32_t i = 0; i < total; ++i) {
 346          if (last_few.size() == WINDOW_SIZE)
 347              last_few.pop_front();
 348          last_few.emplace_back(BLOCK_SIZE, m_rng, set);
 349          uint32_t count = 0;
 350          for (auto& act : last_few)
 351              for (uint32_t k = 0; k < POP_AMOUNT; ++k) {
 352                  count += set.contains(act.reads.back(), true);
 353                  act.reads.pop_back();
 354              }
 355          // We use last_few.size() rather than WINDOW_SIZE for the correct
 356          // behavior on the first WINDOW_SIZE iterations where the deque is not
 357          // full yet.
 358          double hit = (double(count)) / (last_few.size() * POP_AMOUNT);
 359          // Loose Check that hit rate is above min_hit_rate
 360          BOOST_CHECK(hit > min_hit_rate);
 361          // Tighter check, count number of times we are less than tight_hit_rate
 362          // (and implicitly, greater than min_hit_rate)
 363          out_of_tight_tolerance += hit < tight_hit_rate;
 364      }
 365      // Check that being out of tolerance happens less than
 366      // max_rate_less_than_tight_hit_rate of the time
 367      BOOST_CHECK(double(out_of_tight_tolerance) / double(total) < max_rate_less_than_tight_hit_rate);
 368  }
 369  }; // struct GenerationsTest
 370  BOOST_FIXTURE_TEST_CASE(cuckoocache_generations, GenerationsTest)
 371  {
 372      test_cache_generations<CuckooCache::cache<uint256, SignatureCacheHasher>>();
 373  }
 374  
 375  BOOST_AUTO_TEST_SUITE_END();
 376