Block access tracing: Trace referenced key for Get on non-data blocks. (#5548)

Summary:
This PR traces the referenced key for Get for all types of blocks. This is useful when evaluating hybrid row-block caches.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5548

Test Plan: make clean && USE_CLANG=1 make check -j32

Differential Revision: D16157979

Pulled By: HaoyuHuang

fbshipit-source-id: f6327411c9deb74e35e22a35f66cdbae09ab9d87

table/block_based/block_based_table_reader.cc

@@ -1983,10 +1983,12 @@ CachableEntry<UncompressionDict> BlockBasedTable::GetUncompressionDict(
         /*block_size=*/usage, rep_->cf_id_for_tracing(),
         /*cf_name=*/"", rep_->level_for_tracing(),
         rep_->sst_number_for_tracing(), lookup_context->caller, is_cache_hit,
-        /*no_insert=*/no_io, lookup_context->get_id);
+        /*no_insert=*/no_io, lookup_context->get_id,
+        lookup_context->get_from_user_specified_snapshot,
+        /*referenced_key=*/"");
     block_cache_tracer_->WriteBlockAccess(access_record, cache_key,
                                           rep_->cf_name_for_tracing(),
-                                          /*referenced_key=*/nullptr);
+                                          lookup_context->referenced_key);
   }
   return {dict, cache_handle ? rep_->table_options.block_cache.get() : nullptr,
           cache_handle, false /* own_value */};
@@ -2237,7 +2239,6 @@ Status BlockBasedTable::MaybeReadBlockAndLoadToCache(
   Slice key /* key to the block cache */;
   Slice ckey /* key to the compressed block cache */;
   bool is_cache_hit = false;
-  bool no_insert = true;
   if (block_cache != nullptr || block_cache_compressed != nullptr) {
     // create key for block cache
     if (block_cache != nullptr) {
@@ -2265,7 +2266,6 @@ Status BlockBasedTable::MaybeReadBlockAndLoadToCache(
     // Can't find the block from the cache. If I/O is allowed, read from the
     // file.
     if (block_entry->GetValue() == nullptr && !no_io && ro.fill_cache) {
-      no_insert = false;
       Statistics* statistics = rep_->ioptions.statistics;
       const bool maybe_compressed =
           block_type != BlockType::kFilter && rep_->blocks_maybe_compressed;
@@ -2332,11 +2332,11 @@ Status BlockBasedTable::MaybeReadBlockAndLoadToCache(
         assert(false);
         break;
     }
-    if (BlockCacheTraceHelper::ShouldTraceReferencedKey(
+    bool no_insert = no_io || !ro.fill_cache;
+    if (BlockCacheTraceHelper::IsGetOrMultiGetOnDataBlock(
             trace_block_type, lookup_context->caller)) {
       // Defer logging the access to Get() and MultiGet() to trace additional
-      // information, e.g., the referenced key,
+      // information, e.g., referenced_key_exist_in_block.
-      // referenced_key_exist_in_block.
 
       // Make a copy of the block key here since it will be logged later.
       lookup_context->FillLookupContext(
@@ -2351,10 +2351,12 @@ Status BlockBasedTable::MaybeReadBlockAndLoadToCache(
           /*block_size=*/usage, rep_->cf_id_for_tracing(),
           /*cf_name=*/"", rep_->level_for_tracing(),
           rep_->sst_number_for_tracing(), lookup_context->caller, is_cache_hit,
-          no_insert, lookup_context->get_id);
+          no_insert, lookup_context->get_id,
+          lookup_context->get_from_user_specified_snapshot,
+          /*referenced_key=*/"");
       block_cache_tracer_->WriteBlockAccess(access_record, key,
                                             rep_->cf_name_for_tracing(),
-                                            /*referenced_key=*/nullptr);
+                                            lookup_context->referenced_key);
     }
   }
 
@@ -3288,12 +3290,18 @@ Status BlockBasedTable::Get(const ReadOptions& read_options, const Slice& key,
   // First check the full filter
   // If full filter not useful, Then go into each block
   uint64_t tracing_get_id = get_context->get_tracing_get_id();
-  BlockCacheLookupContext lookup_context{TableReaderCaller::kUserGet,
+  BlockCacheLookupContext lookup_context{
-                                         tracing_get_id};
+      TableReaderCaller::kUserGet, tracing_get_id,
+      /*get_from_user_specified_snapshot=*/read_options.snapshot != nullptr};
+  if (block_cache_tracer_ && block_cache_tracer_->is_tracing_enabled()) {
+    // Trace the key since it contains both user key and sequence number.
+    lookup_context.referenced_key = key.ToString();
+    lookup_context.get_from_user_specified_snapshot =
+        read_options.snapshot != nullptr;
+  }
   const bool may_match =
       FullFilterKeyMayMatch(read_options, filter, key, no_io, prefix_extractor,
                             get_context, &lookup_context);
-
   if (!may_match) {
     RecordTick(rep_->ioptions.statistics, BLOOM_FILTER_USEFUL);
     PERF_COUNTER_BY_LEVEL_ADD(bloom_filter_useful, 1, rep_->level);
@@ -3347,7 +3355,9 @@ Status BlockBasedTable::Get(const ReadOptions& read_options, const Slice& key,
       }
 
       BlockCacheLookupContext lookup_data_block_context{
-          TableReaderCaller::kUserGet, tracing_get_id};
+          TableReaderCaller::kUserGet, tracing_get_id,
+          /*get_from_user_specified_snapshot=*/read_options.snapshot !=
+              nullptr};
       bool does_referenced_key_exist = false;
       DataBlockIter biter;
       uint64_t referenced_data_size = 0;
@@ -3406,7 +3416,7 @@ Status BlockBasedTable::Get(const ReadOptions& read_options, const Slice& key,
         if (does_referenced_key_exist) {
           referenced_key = biter.key();
         } else {
-          referenced_key = ExtractUserKey(key);
+          referenced_key = key;
         }
         BlockCacheTraceRecord access_record(
             rep_->ioptions.env->NowMicros(),
@@ -3417,6 +3427,7 @@ Status BlockBasedTable::Get(const ReadOptions& read_options, const Slice& key,
             lookup_data_block_context.is_cache_hit,
             lookup_data_block_context.no_insert,
             lookup_data_block_context.get_id,
+            lookup_data_block_context.get_from_user_specified_snapshot,
             /*referenced_key=*/"", referenced_data_size,
             lookup_data_block_context.num_keys_in_block,
             does_referenced_key_exist);
@@ -3460,8 +3471,9 @@ void BlockBasedTable::MultiGet(const ReadOptions& read_options,
   if (!sst_file_range.empty() && sst_file_range.begin()->get_context) {
     tracing_mget_id = sst_file_range.begin()->get_context->get_tracing_get_id();
   }
-  BlockCacheLookupContext lookup_context{TableReaderCaller::kUserMultiGet,
+  BlockCacheLookupContext lookup_context{
-                                         tracing_mget_id};
+      TableReaderCaller::kUserMultiGet, tracing_mget_id,
+      /*get_from_user_specified_snapshot=*/read_options.snapshot != nullptr};
   FullFilterKeysMayMatch(read_options, filter, &sst_file_range, no_io,
                          prefix_extractor, &lookup_context);
 
@@ -3492,11 +3504,8 @@ void BlockBasedTable::MultiGet(const ReadOptions& read_options,
     {
       MultiGetRange data_block_range(sst_file_range, sst_file_range.begin(),
                                      sst_file_range.end());
-      BlockCacheLookupContext lookup_compression_dict_context(
+      auto uncompression_dict_storage = GetUncompressionDict(
-          TableReaderCaller::kUserMultiGet);
+          nullptr, no_io, sst_file_range.begin()->get_context, &lookup_context);
-      auto uncompression_dict_storage = GetUncompressionDict(nullptr, no_io,
-                                          sst_file_range.begin()->get_context,
-                                          &lookup_compression_dict_context);
       const UncompressionDict& uncompression_dict =
           uncompression_dict_storage.GetValue() == nullptr
               ? UncompressionDict::GetEmptyDict()
@@ -3591,7 +3600,9 @@ void BlockBasedTable::MultiGet(const ReadOptions& read_options,
         uint64_t referenced_data_size = 0;
         bool does_referenced_key_exist = false;
         BlockCacheLookupContext lookup_data_block_context(
-            TableReaderCaller::kUserMultiGet, tracing_mget_id);
+            TableReaderCaller::kUserMultiGet, tracing_mget_id,
+            /*get_from_user_specified_snapshot=*/read_options.snapshot !=
+                nullptr);
         if (first_block) {
           if (!block_handles[idx_in_batch].IsNull() ||
               !results[idx_in_batch].IsEmpty()) {
@@ -3685,7 +3696,7 @@ void BlockBasedTable::MultiGet(const ReadOptions& read_options,
           if (does_referenced_key_exist) {
             referenced_key = biter->key();
           } else {
-            referenced_key = ExtractUserKey(key);
+            referenced_key = key;
           }
           BlockCacheTraceRecord access_record(
               rep_->ioptions.env->NowMicros(),
@@ -3696,6 +3707,7 @@ void BlockBasedTable::MultiGet(const ReadOptions& read_options,
               lookup_data_block_context.is_cache_hit,
               lookup_data_block_context.no_insert,
               lookup_data_block_context.get_id,
+              lookup_data_block_context.get_from_user_specified_snapshot,
               /*referenced_key=*/"", referenced_data_size,
               lookup_data_block_context.num_keys_in_block,
               does_referenced_key_exist);

table/table_test.cc

@@ -63,6 +63,8 @@ extern const uint64_t kPlainTableMagicNumber;
 
 namespace {
 
+const std::string kDummyValue(10000, 'o');
+
 // DummyPropertiesCollector used to test BlockBasedTableProperties
 class DummyPropertiesCollector : public TablePropertiesCollector {
  public:
@@ -312,7 +314,9 @@ class TableConstructor: public Constructor {
       : Constructor(cmp),
         largest_seqno_(largest_seqno),
         convert_to_internal_key_(convert_to_internal_key),
-        level_(level) {}
+        level_(level) {
+    env_ = rocksdb::Env::Default();
+  }
   ~TableConstructor() override { Reset(); }
 
   Status FinishImpl(const Options& options, const ImmutableCFOptions& ioptions,
@@ -371,7 +375,7 @@ class TableConstructor: public Constructor {
     return ioptions.table_factory->NewTableReader(
         TableReaderOptions(ioptions, moptions.prefix_extractor.get(), soptions,
                            internal_comparator, !kSkipFilters, !kImmortal,
-                           level_, largest_seqno_, nullptr),
+                           level_, largest_seqno_, &block_cache_tracer_),
         std::move(file_reader_), TEST_GetSink()->contents().size(),
         &table_reader_);
   }
@@ -425,6 +429,8 @@ class TableConstructor: public Constructor {
     return static_cast<test::StringSink*>(file_writer_->writable_file());
   }
 
+  BlockCacheTracer block_cache_tracer_;
+
  private:
   void Reset() {
     uniq_id_ = 0;
@@ -445,6 +451,7 @@ class TableConstructor: public Constructor {
 
   static uint64_t cur_uniq_id_;
   EnvOptions soptions;
+  Env* env_;
 };
 uint64_t TableConstructor::cur_uniq_id_ = 1;
 
@@ -1063,7 +1070,9 @@ class BlockBasedTableTest
     : public TableTest,
       virtual public ::testing::WithParamInterface<uint32_t> {
  public:
-  BlockBasedTableTest() : format_(GetParam()) {}
+  BlockBasedTableTest() : format_(GetParam()) {
+    env_ = rocksdb::Env::Default();
+  }
 
   BlockBasedTableOptions GetBlockBasedTableOptions() {
     BlockBasedTableOptions options;
@@ -1071,11 +1080,91 @@ class BlockBasedTableTest
     return options;
   }
 
+  void SetupTracingTest(TableConstructor* c) {
+    test_path_ = test::PerThreadDBPath("block_based_table_tracing_test");
+    EXPECT_OK(env_->CreateDir(test_path_));
+    trace_file_path_ = test_path_ + "/block_cache_trace_file";
+    TraceOptions trace_opt;
+    std::unique_ptr<TraceWriter> trace_writer;
+    EXPECT_OK(NewFileTraceWriter(env_, EnvOptions(), trace_file_path_,
+                                 &trace_writer));
+    c->block_cache_tracer_.StartTrace(env_, trace_opt, std::move(trace_writer));
+    {
+      std::string user_key = "k01";
+      InternalKey internal_key(user_key, 0, kTypeValue);
+      std::string encoded_key = internal_key.Encode().ToString();
+      c->Add(encoded_key, kDummyValue);
+    }
+    {
+      std::string user_key = "k02";
+      InternalKey internal_key(user_key, 0, kTypeValue);
+      std::string encoded_key = internal_key.Encode().ToString();
+      c->Add(encoded_key, kDummyValue);
+    }
+  }
+
+  void VerifyBlockAccessTrace(
+      TableConstructor* c,
+      const std::vector<BlockCacheTraceRecord>& expected_records) {
+    c->block_cache_tracer_.EndTrace();
+
+    std::unique_ptr<TraceReader> trace_reader;
+    Status s =
+        NewFileTraceReader(env_, EnvOptions(), trace_file_path_, &trace_reader);
+    EXPECT_OK(s);
+    BlockCacheTraceReader reader(std::move(trace_reader));
+    BlockCacheTraceHeader header;
+    EXPECT_OK(reader.ReadHeader(&header));
+    uint32_t index = 0;
+    while (s.ok()) {
+      BlockCacheTraceRecord access;
+      s = reader.ReadAccess(&access);
+      if (!s.ok()) {
+        break;
+      }
+      ASSERT_LT(index, expected_records.size());
+      EXPECT_NE("", access.block_key);
+      EXPECT_EQ(access.block_type, expected_records[index].block_type);
+      EXPECT_GT(access.block_size, 0);
+      EXPECT_EQ(access.caller, expected_records[index].caller);
+      EXPECT_EQ(access.no_insert, expected_records[index].no_insert);
+      EXPECT_EQ(access.is_cache_hit, expected_records[index].is_cache_hit);
+      // Get
+      if (access.caller == TableReaderCaller::kUserGet) {
+        EXPECT_EQ(access.referenced_key,
+                  expected_records[index].referenced_key);
+        EXPECT_EQ(access.get_id, expected_records[index].get_id);
+        EXPECT_EQ(access.get_from_user_specified_snapshot,
+                  expected_records[index].get_from_user_specified_snapshot);
+        if (access.block_type == TraceType::kBlockTraceDataBlock) {
+          EXPECT_GT(access.referenced_data_size, 0);
+          EXPECT_GT(access.num_keys_in_block, 0);
+          EXPECT_EQ(access.referenced_key_exist_in_block,
+                    expected_records[index].referenced_key_exist_in_block);
+        }
+      } else {
+        EXPECT_EQ(access.referenced_key, "");
+        EXPECT_EQ(access.get_id, 0);
+        EXPECT_TRUE(access.get_from_user_specified_snapshot == Boolean::kFalse);
+        EXPECT_EQ(access.referenced_data_size, 0);
+        EXPECT_EQ(access.num_keys_in_block, 0);
+        EXPECT_TRUE(access.referenced_key_exist_in_block == Boolean::kFalse);
+      }
+      index++;
+    }
+    EXPECT_EQ(index, expected_records.size());
+    EXPECT_OK(env_->DeleteFile(trace_file_path_));
+    EXPECT_OK(env_->DeleteDir(test_path_));
+  }
+
  protected:
   uint64_t IndexUncompressedHelper(bool indexCompress);
 
  private:
   uint32_t format_;
+  Env* env_;
+  std::string trace_file_path_;
+  std::string test_path_;
 };
 class PlainTableTest : public TableTest {};
 class TablePropertyTest : public testing::Test {};
@@ -2211,6 +2300,187 @@ TEST_P(BlockBasedTableTest, NumBlockStat) {
   c.ResetTableReader();
 }
 
+TEST_P(BlockBasedTableTest, TracingGetTest) {
+  TableConstructor c(BytewiseComparator());
+  Options options;
+  BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
+  options.create_if_missing = true;
+  table_options.block_cache = NewLRUCache(1024 * 1024, 0);
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10, true));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  SetupTracingTest(&c);
+  std::vector<std::string> keys;
+  stl_wrappers::KVMap kvmap;
+  ImmutableCFOptions ioptions(options);
+  MutableCFOptions moptions(options);
+  c.Finish(options, ioptions, moptions, table_options,
+           GetPlainInternalComparator(options.comparator), &keys, &kvmap);
+  std::string user_key = "k01";
+  InternalKey internal_key(user_key, 0, kTypeValue);
+  std::string encoded_key = internal_key.Encode().ToString();
+  for (uint32_t i = 1; i <= 2; i++) {
+    PinnableSlice value;
+    GetContext get_context(options.comparator, nullptr, nullptr, nullptr,
+                           GetContext::kNotFound, user_key, &value, nullptr,
+                           nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+                           nullptr, /*get_id=*/i);
+    get_perf_context()->Reset();
+    ASSERT_OK(c.GetTableReader()->Get(ReadOptions(), encoded_key, &get_context,
+                                      moptions.prefix_extractor.get()));
+    ASSERT_EQ(get_context.State(), GetContext::kFound);
+    ASSERT_EQ(value.ToString(), kDummyValue);
+  }
+
+  // Verify traces.
+  std::vector<BlockCacheTraceRecord> expected_records;
+  // The first two records should be prefetching index and filter blocks.
+  BlockCacheTraceRecord record;
+  record.block_type = TraceType::kBlockTraceIndexBlock;
+  record.caller = TableReaderCaller::kPrefetch;
+  record.is_cache_hit = Boolean::kFalse;
+  record.no_insert = Boolean::kFalse;
+  expected_records.push_back(record);
+  record.block_type = TraceType::kBlockTraceFilterBlock;
+  expected_records.push_back(record);
+  // Then we should have three records for one index, one filter, and one data
+  // block access.
+  record.get_id = 1;
+  record.block_type = TraceType::kBlockTraceIndexBlock;
+  record.caller = TableReaderCaller::kUserGet;
+  record.get_from_user_specified_snapshot = Boolean::kFalse;
+  record.referenced_key = encoded_key;
+  record.referenced_key_exist_in_block = Boolean::kTrue;
+  record.is_cache_hit = Boolean::kTrue;
+  expected_records.push_back(record);
+  record.block_type = TraceType::kBlockTraceFilterBlock;
+  expected_records.push_back(record);
+  record.is_cache_hit = Boolean::kFalse;
+  record.block_type = TraceType::kBlockTraceDataBlock;
+  expected_records.push_back(record);
+  // The second get should all observe cache hits.
+  record.is_cache_hit = Boolean::kTrue;
+  record.get_id = 2;
+  record.block_type = TraceType::kBlockTraceIndexBlock;
+  record.caller = TableReaderCaller::kUserGet;
+  record.get_from_user_specified_snapshot = Boolean::kFalse;
+  record.referenced_key = encoded_key;
+  expected_records.push_back(record);
+  record.block_type = TraceType::kBlockTraceFilterBlock;
+  expected_records.push_back(record);
+  record.block_type = TraceType::kBlockTraceDataBlock;
+  expected_records.push_back(record);
+  VerifyBlockAccessTrace(&c, expected_records);
+  c.ResetTableReader();
+}
+
+TEST_P(BlockBasedTableTest, TracingApproximateOffsetOfTest) {
+  TableConstructor c(BytewiseComparator());
+  Options options;
+  BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
+  options.create_if_missing = true;
+  table_options.block_cache = NewLRUCache(1024 * 1024, 0);
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10, true));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  SetupTracingTest(&c);
+  std::vector<std::string> keys;
+  stl_wrappers::KVMap kvmap;
+  ImmutableCFOptions ioptions(options);
+  MutableCFOptions moptions(options);
+  c.Finish(options, ioptions, moptions, table_options,
+           GetPlainInternalComparator(options.comparator), &keys, &kvmap);
+  for (uint32_t i = 1; i <= 2; i++) {
+    std::string user_key = "k01";
+    InternalKey internal_key(user_key, 0, kTypeValue);
+    std::string encoded_key = internal_key.Encode().ToString();
+    c.GetTableReader()->ApproximateOffsetOf(
+        encoded_key, TableReaderCaller::kUserApproximateSize);
+  }
+  // Verify traces.
+  std::vector<BlockCacheTraceRecord> expected_records;
+  // The first two records should be prefetching index and filter blocks.
+  BlockCacheTraceRecord record;
+  record.block_type = TraceType::kBlockTraceIndexBlock;
+  record.caller = TableReaderCaller::kPrefetch;
+  record.is_cache_hit = Boolean::kFalse;
+  record.no_insert = Boolean::kFalse;
+  expected_records.push_back(record);
+  record.block_type = TraceType::kBlockTraceFilterBlock;
+  expected_records.push_back(record);
+  // Then we should have two records for only index blocks.
+  record.block_type = TraceType::kBlockTraceIndexBlock;
+  record.caller = TableReaderCaller::kUserApproximateSize;
+  record.is_cache_hit = Boolean::kTrue;
+  expected_records.push_back(record);
+  expected_records.push_back(record);
+  VerifyBlockAccessTrace(&c, expected_records);
+  c.ResetTableReader();
+}
+
+TEST_P(BlockBasedTableTest, TracingIterator) {
+  TableConstructor c(BytewiseComparator());
+  Options options;
+  BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
+  options.create_if_missing = true;
+  table_options.block_cache = NewLRUCache(1024 * 1024, 0);
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10, true));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  SetupTracingTest(&c);
+  std::vector<std::string> keys;
+  stl_wrappers::KVMap kvmap;
+  ImmutableCFOptions ioptions(options);
+  MutableCFOptions moptions(options);
+  c.Finish(options, ioptions, moptions, table_options,
+           GetPlainInternalComparator(options.comparator), &keys, &kvmap);
+
+  for (uint32_t i = 1; i <= 2; i++) {
+    std::unique_ptr<InternalIterator> iter(c.GetTableReader()->NewIterator(
+        ReadOptions(), moptions.prefix_extractor.get(), /*arena=*/nullptr,
+        /*skip_filters=*/false, TableReaderCaller::kUserIterator));
+    iter->SeekToFirst();
+    while (iter->Valid()) {
+      iter->key();
+      iter->value();
+      iter->Next();
+    }
+    ASSERT_OK(iter->status());
+    iter.reset();
+  }
+
+  // Verify traces.
+  std::vector<BlockCacheTraceRecord> expected_records;
+  // The first two records should be prefetching index and filter blocks.
+  BlockCacheTraceRecord record;
+  record.block_type = TraceType::kBlockTraceIndexBlock;
+  record.caller = TableReaderCaller::kPrefetch;
+  record.is_cache_hit = Boolean::kFalse;
+  record.no_insert = Boolean::kFalse;
+  expected_records.push_back(record);
+  record.block_type = TraceType::kBlockTraceFilterBlock;
+  expected_records.push_back(record);
+  // Then we should have three records for index and two data block access.
+  record.block_type = TraceType::kBlockTraceIndexBlock;
+  record.caller = TableReaderCaller::kUserIterator;
+  record.is_cache_hit = Boolean::kTrue;
+  expected_records.push_back(record);
+  record.block_type = TraceType::kBlockTraceDataBlock;
+  record.is_cache_hit = Boolean::kFalse;
+  expected_records.push_back(record);
+  expected_records.push_back(record);
+  // When we iterate this file for the second time, we should observe all cache
+  // hits.
+  record.block_type = TraceType::kBlockTraceIndexBlock;
+  record.is_cache_hit = Boolean::kTrue;
+  expected_records.push_back(record);
+  record.block_type = TraceType::kBlockTraceDataBlock;
+  expected_records.push_back(record);
+  expected_records.push_back(record);
+  VerifyBlockAccessTrace(&c, expected_records);
+  c.ResetTableReader();
+}
+
 // A simple tool that takes the snapshot of block cache statistics.
 class BlockCachePropertiesSnapshot {
  public:

tools/block_cache_trace_analyzer.h

@@ -57,7 +57,7 @@ struct BlockAccessInfo {
     const uint64_t timestamp_in_seconds =
         access.access_timestamp / kMicrosInSecond;
     caller_num_accesses_timeline[access.caller][timestamp_in_seconds] += 1;
-    if (BlockCacheTraceHelper::ShouldTraceReferencedKey(access.block_type,
+    if (BlockCacheTraceHelper::IsGetOrMultiGetOnDataBlock(access.block_type,
                                                           access.caller)) {
       num_keys = access.num_keys_in_block;
       if (access.referenced_key_exist_in_block == Boolean::kTrue) {

trace_replay/block_cache_tracer.cc

@@ -35,14 +35,13 @@ const std::string BlockCacheTraceHelper::kUnknownColumnFamilyName =
     "UnknownColumnFamily";
 const uint64_t BlockCacheTraceHelper::kReservedGetId = 0;
 
-bool BlockCacheTraceHelper::ShouldTraceReferencedKey(TraceType block_type,
+bool BlockCacheTraceHelper::IsGetOrMultiGetOnDataBlock(
-                                                     TableReaderCaller caller) {
+    TraceType block_type, TableReaderCaller caller) {
   return (block_type == TraceType::kBlockTraceDataBlock) &&
-         (caller == TableReaderCaller::kUserGet ||
+         IsGetOrMultiGet(caller);
-          caller == TableReaderCaller::kUserMultiGet);
 }
 
-bool BlockCacheTraceHelper::ShouldTraceGetId(TableReaderCaller caller) {
+bool BlockCacheTraceHelper::IsGetOrMultiGet(TableReaderCaller caller) {
   return caller == TableReaderCaller::kUserGet ||
          caller == TableReaderCaller::kUserMultiGet;
 }
@@ -81,12 +80,13 @@ Status BlockCacheTraceWriter::WriteBlockAccess(
   trace.payload.push_back(record.caller);
   trace.payload.push_back(record.is_cache_hit);
   trace.payload.push_back(record.no_insert);
-  if (BlockCacheTraceHelper::ShouldTraceGetId(record.caller)) {
+  if (BlockCacheTraceHelper::IsGetOrMultiGet(record.caller)) {
     PutFixed64(&trace.payload, record.get_id);
+    trace.payload.push_back(record.get_from_user_specified_snapshot);
+    PutLengthPrefixedSlice(&trace.payload, referenced_key);
   }
-  if (BlockCacheTraceHelper::ShouldTraceReferencedKey(record.block_type,
+  if (BlockCacheTraceHelper::IsGetOrMultiGetOnDataBlock(record.block_type,
                                                         record.caller)) {
-    PutLengthPrefixedSlice(&trace.payload, referenced_key);
     PutFixed64(&trace.payload, record.referenced_data_size);
     PutFixed64(&trace.payload, record.num_keys_in_block);
     trace.payload.push_back(record.referenced_key_exist_in_block);
@@ -216,20 +216,28 @@ Status BlockCacheTraceReader::ReadAccess(BlockCacheTraceRecord* record) {
   }
   record->no_insert = static_cast<Boolean>(enc_slice[0]);
   enc_slice.remove_prefix(kCharSize);
-  if (BlockCacheTraceHelper::ShouldTraceGetId(record->caller)) {
+  if (BlockCacheTraceHelper::IsGetOrMultiGet(record->caller)) {
     if (!GetFixed64(&enc_slice, &record->get_id)) {
       return Status::Incomplete(
           "Incomplete access record: Failed to read the get id.");
     }
+    if (enc_slice.empty()) {
+      return Status::Incomplete(
+          "Incomplete access record: Failed to read "
+          "get_from_user_specified_snapshot.");
     }
-  if (BlockCacheTraceHelper::ShouldTraceReferencedKey(record->block_type,
+    record->get_from_user_specified_snapshot =
-                                                      record->caller)) {
+        static_cast<Boolean>(enc_slice[0]);
+    enc_slice.remove_prefix(kCharSize);
     Slice referenced_key;
     if (!GetLengthPrefixedSlice(&enc_slice, &referenced_key)) {
       return Status::Incomplete(
           "Incomplete access record: Failed to read the referenced key.");
     }
     record->referenced_key = referenced_key.ToString();
+  }
+  if (BlockCacheTraceHelper::IsGetOrMultiGetOnDataBlock(record->block_type,
+                                                        record->caller)) {
     if (!GetFixed64(&enc_slice, &record->referenced_data_size)) {
       return Status::Incomplete(
           "Incomplete access record: Failed to read the referenced data size.");

trace_replay/block_cache_tracer.h

@@ -23,9 +23,9 @@ extern const uint64_t kSecondInHour;
 
 class BlockCacheTraceHelper {
  public:
-  static bool ShouldTraceReferencedKey(TraceType block_type,
+  static bool IsGetOrMultiGetOnDataBlock(TraceType block_type,
                                          TableReaderCaller caller);
-  static bool ShouldTraceGetId(TableReaderCaller caller);
+  static bool IsGetOrMultiGet(TableReaderCaller caller);
   static bool IsUserAccess(TableReaderCaller caller);
 
   static const std::string kUnknownColumnFamilyName;
@@ -53,8 +53,11 @@ class BlockCacheTraceHelper {
 // kUserApproximateSize).
 struct BlockCacheLookupContext {
   BlockCacheLookupContext(const TableReaderCaller& _caller) : caller(_caller) {}
-  BlockCacheLookupContext(const TableReaderCaller& _caller, uint64_t _get_id)
+  BlockCacheLookupContext(const TableReaderCaller& _caller, uint64_t _get_id,
-      : caller(_caller), get_id(_get_id) {}
+                          bool _get_from_user_specified_snapshot)
+      : caller(_caller),
+        get_id(_get_id),
+        get_from_user_specified_snapshot(_get_from_user_specified_snapshot) {}
   const TableReaderCaller caller;
   // These are populated when we perform lookup/insert on block cache. The block
   // cache tracer uses these inforation when logging the block access at
@@ -69,6 +72,8 @@ struct BlockCacheLookupContext {
   // how many blocks a Get/MultiGet request accesses. We can also measure the
   // impact of row cache vs block cache.
   uint64_t get_id = 0;
+  std::string referenced_key;
+  bool get_from_user_specified_snapshot = false;
 
   void FillLookupContext(bool _is_cache_hit, bool _no_insert,
                          TraceType _block_type, uint64_t _block_size,
@@ -100,21 +105,23 @@ struct BlockCacheTraceRecord {
   Boolean no_insert = Boolean::kFalse;
   // Required field for Get and MultiGet
   uint64_t get_id = BlockCacheTraceHelper::kReservedGetId;
-  // Required fields for data block and user Get/Multi-Get only.
+  Boolean get_from_user_specified_snapshot = Boolean::kFalse;
   std::string referenced_key;
+  // Required fields for data block and user Get/Multi-Get only.
   uint64_t referenced_data_size = 0;
   uint64_t num_keys_in_block = 0;
   Boolean referenced_key_exist_in_block = Boolean::kFalse;
 
   BlockCacheTraceRecord() {}
 
-  BlockCacheTraceRecord(uint64_t _access_timestamp, std::string _block_key,
+  BlockCacheTraceRecord(
-                        TraceType _block_type, uint64_t _block_size,
+      uint64_t _access_timestamp, std::string _block_key, TraceType _block_type,
-                        uint64_t _cf_id, std::string _cf_name, uint32_t _level,
+      uint64_t _block_size, uint64_t _cf_id, std::string _cf_name,
-                        uint64_t _sst_fd_number, TableReaderCaller _caller,
+      uint32_t _level, uint64_t _sst_fd_number, TableReaderCaller _caller,
-                        bool _is_cache_hit, bool _no_insert, uint64_t _get_id,
+      bool _is_cache_hit, bool _no_insert,
-                        std::string _referenced_key = "",
+      uint64_t _get_id = BlockCacheTraceHelper::kReservedGetId,
-                        uint64_t _referenced_data_size = 0,
+      bool _get_from_user_specified_snapshot = false,
+      std::string _referenced_key = "", uint64_t _referenced_data_size = 0,
       uint64_t _num_keys_in_block = 0,
       bool _referenced_key_exist_in_block = false)
       : access_timestamp(_access_timestamp),
@@ -129,6 +136,9 @@ struct BlockCacheTraceRecord {
         is_cache_hit(_is_cache_hit ? Boolean::kTrue : Boolean::kFalse),
         no_insert(_no_insert ? Boolean::kTrue : Boolean::kFalse),
         get_id(_get_id),
+        get_from_user_specified_snapshot(_get_from_user_specified_snapshot
+                                             ? Boolean::kTrue
+                                             : Boolean::kFalse),
         referenced_key(_referenced_key),
         referenced_data_size(_referenced_data_size),
         num_keys_in_block(_num_keys_in_block),

trace_replay/block_cache_tracer_test.cc

@@ -74,6 +74,7 @@ class BlockCacheTracerTest : public testing::Test {
       // Provide get_id for all callers. The writer should only write get_id
       // when the caller is either GET or MGET.
       record.get_id = key_id + 1;
+      record.get_from_user_specified_snapshot = Boolean::kTrue;
       // Provide these fields for all block types.
       // The writer should only write these fields for data blocks and the
       // caller is either GET or MGET.
@@ -126,20 +127,22 @@ class BlockCacheTracerTest : public testing::Test {
       if (record.caller == TableReaderCaller::kUserGet ||
           record.caller == TableReaderCaller::kUserMultiGet) {
         ASSERT_EQ(key_id + 1, record.get_id);
+        ASSERT_EQ(Boolean::kTrue, record.get_from_user_specified_snapshot);
+        ASSERT_EQ(kRefKeyPrefix + std::to_string(key_id),
+                  record.referenced_key);
       } else {
         ASSERT_EQ(BlockCacheTraceHelper::kReservedGetId, record.get_id);
+        ASSERT_EQ(Boolean::kFalse, record.get_from_user_specified_snapshot);
+        ASSERT_EQ("", record.referenced_key);
       }
       if (block_type == TraceType::kBlockTraceDataBlock &&
           (record.caller == TableReaderCaller::kUserGet ||
            record.caller == TableReaderCaller::kUserMultiGet)) {
-        ASSERT_EQ(kRefKeyPrefix + std::to_string(key_id),
-                  record.referenced_key);
         ASSERT_EQ(Boolean::kTrue, record.referenced_key_exist_in_block);
         ASSERT_EQ(kNumKeysInBlock, record.num_keys_in_block);
         ASSERT_EQ(kReferencedDataSize + key_id, record.referenced_data_size);
         continue;
       }
-      ASSERT_EQ("", record.referenced_key);
       ASSERT_EQ(Boolean::kFalse, record.referenced_key_exist_in_block);
       ASSERT_EQ(0, record.num_keys_in_block);
       ASSERT_EQ(0, record.referenced_data_size);

utilities/simulator_cache/cache_simulator.cc

@@ -110,19 +110,22 @@ void PrioritizedCacheSimulator::Access(const BlockCacheTraceRecord& access) {
 std::string HybridRowBlockCacheSimulator::ComputeRowKey(
     const BlockCacheTraceRecord& access) {
   assert(access.get_id != BlockCacheTraceHelper::kReservedGetId);
-  Slice key;
+  Slice key = ExtractUserKey(access.referenced_key);
-  if (access.referenced_key_exist_in_block == Boolean::kTrue) {
+  uint64_t seq_no = access.get_from_user_specified_snapshot == Boolean::kFalse
-    key = ExtractUserKey(access.referenced_key);
+                        ? 0
-  } else {
+                        : 1 + GetInternalKeySeqno(access.referenced_key);
-    key = access.referenced_key;
+  return std::to_string(access.sst_fd_number) + "_" + key.ToString() + "_" +
-  }
+         std::to_string(seq_no);
-  return std::to_string(access.sst_fd_number) + "_" + key.ToString();
 }
 
 void HybridRowBlockCacheSimulator::Access(const BlockCacheTraceRecord& access) {
   bool is_cache_miss = true;
   bool admitted = true;
-  if (access.get_id != BlockCacheTraceHelper::kReservedGetId) {
+  // TODO (haoyu): We only support Get for now. We need to extend the tracing
+  // for MultiGet, i.e., non-data block accesses must log all keys in a
+  // MultiGet.
+  if (access.caller == TableReaderCaller::kUserGet &&
+      access.get_id != BlockCacheTraceHelper::kReservedGetId) {
     // This is a Get/MultiGet request.
     const std::string& row_key = ComputeRowKey(access);
     if (getid_getkeys_map_[access.get_id].find(row_key) ==

utilities/simulator_cache/cache_simulator.h

@@ -137,7 +137,6 @@ class HybridRowBlockCacheSimulator : public PrioritizedCacheSimulator {
  private:
   // Row key is a concatenation of the access's fd_number and the referenced
   // user key.
-  // TODO(haoyu): the row key should contain sequence number.
   std::string ComputeRowKey(const BlockCacheTraceRecord& access);
 
   enum InsertResult : char {

utilities/simulator_cache/cache_simulator_test.cc

@@ -174,10 +174,11 @@ TEST_F(CacheSimulatorTest, GhostPrioritizedCacheSimulator) {
 TEST_F(CacheSimulatorTest, HybridRowBlockCacheSimulator) {
   uint64_t block_id = 100;
   BlockCacheTraceRecord first_get = GenerateGetRecord(kGetId);
+  first_get.get_from_user_specified_snapshot = Boolean::kTrue;
   BlockCacheTraceRecord second_get = GenerateGetRecord(kGetId + 1);
   second_get.referenced_data_size = 0;
   second_get.referenced_key_exist_in_block = Boolean::kFalse;
-  second_get.referenced_key = kRefKeyPrefix + std::to_string(kGetId);
+  second_get.get_from_user_specified_snapshot = Boolean::kTrue;
   BlockCacheTraceRecord third_get = GenerateGetRecord(kGetId + 2);
   third_get.referenced_data_size = 0;
   third_get.referenced_key_exist_in_block = Boolean::kFalse;
@@ -203,9 +204,10 @@ TEST_F(CacheSimulatorTest, HybridRowBlockCacheSimulator) {
   ASSERT_EQ(100, cache_simulator->miss_ratio());
   ASSERT_EQ(10, cache_simulator->user_accesses());
   ASSERT_EQ(100, cache_simulator->user_miss_ratio());
-  auto handle =
+  auto handle = sim_cache->Lookup(
-      sim_cache->Lookup(ExtractUserKey(std::to_string(first_get.sst_fd_number) +
+      std::to_string(first_get.sst_fd_number) + "_" +
-                                       "_" + first_get.referenced_key));
+      ExtractUserKey(first_get.referenced_key).ToString() + "_" +
+      std::to_string(1 + GetInternalKeySeqno(first_get.referenced_key)));
   ASSERT_NE(nullptr, handle);
   sim_cache->Release(handle);
   for (uint32_t i = 100; i < block_id; i++) {
@@ -227,8 +229,10 @@ TEST_F(CacheSimulatorTest, HybridRowBlockCacheSimulator) {
   ASSERT_EQ(66, static_cast<uint64_t>(cache_simulator->miss_ratio()));
   ASSERT_EQ(15, cache_simulator->user_accesses());
   ASSERT_EQ(66, static_cast<uint64_t>(cache_simulator->user_miss_ratio()));
-  handle = sim_cache->Lookup(std::to_string(second_get.sst_fd_number) + "_" +
+  handle = sim_cache->Lookup(
-                             second_get.referenced_key);
+      std::to_string(second_get.sst_fd_number) + "_" +
+      ExtractUserKey(second_get.referenced_key).ToString() + "_" +
+      std::to_string(1 + GetInternalKeySeqno(second_get.referenced_key)));
   ASSERT_NE(nullptr, handle);
   sim_cache->Release(handle);
   for (uint32_t i = 100; i < block_id; i++) {
@@ -283,9 +287,9 @@ TEST_F(CacheSimulatorTest, HybridRowBlockNoInsertCacheSimulator) {
     cache_simulator->Access(first_get);
     block_id++;
   }
-  auto handle =
+  auto handle = sim_cache->Lookup(
-      sim_cache->Lookup(ExtractUserKey(std::to_string(first_get.sst_fd_number) +
+      std::to_string(first_get.sst_fd_number) + "_" +
-                                       "_" + first_get.referenced_key));
+      ExtractUserKey(first_get.referenced_key).ToString() + "_0");
   ASSERT_NE(nullptr, handle);
   sim_cache->Release(handle);
   // All blocks are missing from the cache since insert_blocks_row_kvpair_misses