diff --git a/db/perf_context_test.cc b/db/perf_context_test.cc
index 9494ac92b..8a345e5bb 100644
--- a/db/perf_context_test.cc
+++ b/db/perf_context_test.cc
@@ -589,6 +589,19 @@ TEST_F(PerfContextTest, FalseDBMutexWait) {
     }
   }
 }
+
+TEST_F(PerfContextTest, ToString) {
+  perf_context.Reset();
+  perf_context.block_read_count = 12345;
+
+  std::string zero_included = perf_context.ToString();
+  ASSERT_NE(std::string::npos, zero_included.find("= 0"));
+  ASSERT_NE(std::string::npos, zero_included.find("= 12345"));
+
+  std::string zero_excluded = perf_context.ToString(true);
+  ASSERT_EQ(std::string::npos, zero_excluded.find("= 0"));
+  ASSERT_NE(std::string::npos, zero_excluded.find("= 12345"));
+}
 }
 
 int main(int argc, char** argv) {
diff --git a/include/rocksdb/perf_context.h b/include/rocksdb/perf_context.h
index c2af729e3..7a6b6f367 100644
--- a/include/rocksdb/perf_context.h
+++ b/include/rocksdb/perf_context.h
@@ -21,7 +21,7 @@ struct PerfContext {
 
   void Reset(); // reset all performance counters to zero
 
-  std::string ToString() const;
+  std::string ToString(bool exclude_zero_counters = false) const;
 
   uint64_t user_key_comparison_count; // total number of user key comparisons
   uint64_t block_cache_hit_count;     // total number of block cache hits
diff --git a/util/perf_context.cc b/util/perf_context.cc
index 282516590..07bad40f2 100644
--- a/util/perf_context.cc
+++ b/util/perf_context.cc
@@ -61,32 +61,54 @@ void PerfContext::Reset() {
 #endif
 }
 
-#define OUTPUT(counter) #counter << " = " << counter << ", "
+#define OUTPUT(counter)                          \
+  if (!exclude_zero_counters || (counter > 0)) { \
+    ss << #counter << " = " << counter << ", ";  \
+  }
 
-std::string PerfContext::ToString() const {
+std::string PerfContext::ToString(bool exclude_zero_counters) const {
 #if defined(NPERF_CONTEXT) || defined(IOS_CROSS_COMPILE)
   return "";
 #else
   std::ostringstream ss;
-  ss << OUTPUT(user_key_comparison_count) << OUTPUT(block_cache_hit_count)
-     << OUTPUT(block_read_count) << OUTPUT(block_read_byte)
-     << OUTPUT(block_read_time) << OUTPUT(block_checksum_time)
-     << OUTPUT(block_decompress_time) << OUTPUT(internal_key_skipped_count)
-     << OUTPUT(internal_delete_skipped_count) << OUTPUT(write_wal_time)
-     << OUTPUT(get_snapshot_time) << OUTPUT(get_from_memtable_time)
-     << OUTPUT(get_from_memtable_count) << OUTPUT(get_post_process_time)
-     << OUTPUT(get_from_output_files_time) << OUTPUT(seek_on_memtable_time)
-     << OUTPUT(seek_on_memtable_count) << OUTPUT(seek_child_seek_time)
-     << OUTPUT(seek_child_seek_count) << OUTPUT(seek_min_heap_time)
-     << OUTPUT(seek_internal_seek_time) << OUTPUT(find_next_user_entry_time)
-     << OUTPUT(write_pre_and_post_process_time) << OUTPUT(write_memtable_time)
-     << OUTPUT(db_mutex_lock_nanos) << OUTPUT(db_condition_wait_nanos)
-     << OUTPUT(merge_operator_time_nanos) << OUTPUT(write_delay_time)
-     << OUTPUT(read_index_block_nanos) << OUTPUT(read_filter_block_nanos)
-     << OUTPUT(new_table_block_iter_nanos) << OUTPUT(new_table_iterator_nanos)
-     << OUTPUT(block_seek_nanos) << OUTPUT(find_table_nanos)
-     << OUTPUT(bloom_memtable_hit_count) << OUTPUT(bloom_memtable_miss_count)
-     << OUTPUT(bloom_sst_hit_count) << OUTPUT(bloom_sst_miss_count);
+  OUTPUT(user_key_comparison_count);
+  OUTPUT(block_cache_hit_count);
+  OUTPUT(block_read_count);
+  OUTPUT(block_read_byte);
+  OUTPUT(block_read_time);
+  OUTPUT(block_checksum_time);
+  OUTPUT(block_decompress_time);
+  OUTPUT(internal_key_skipped_count);
+  OUTPUT(internal_delete_skipped_count);
+  OUTPUT(write_wal_time);
+  OUTPUT(get_snapshot_time);
+  OUTPUT(get_from_memtable_time);
+  OUTPUT(get_from_memtable_count);
+  OUTPUT(get_post_process_time);
+  OUTPUT(get_from_output_files_time);
+  OUTPUT(seek_on_memtable_time);
+  OUTPUT(seek_on_memtable_count);
+  OUTPUT(seek_child_seek_time);
+  OUTPUT(seek_child_seek_count);
+  OUTPUT(seek_min_heap_time);
+  OUTPUT(seek_internal_seek_time);
+  OUTPUT(find_next_user_entry_time);
+  OUTPUT(write_pre_and_post_process_time);
+  OUTPUT(write_memtable_time);
+  OUTPUT(db_mutex_lock_nanos);
+  OUTPUT(db_condition_wait_nanos);
+  OUTPUT(merge_operator_time_nanos);
+  OUTPUT(write_delay_time);
+  OUTPUT(read_index_block_nanos);
+  OUTPUT(read_filter_block_nanos);
+  OUTPUT(new_table_block_iter_nanos);
+  OUTPUT(new_table_iterator_nanos);
+  OUTPUT(block_seek_nanos);
+  OUTPUT(find_table_nanos);
+  OUTPUT(bloom_memtable_hit_count);
+  OUTPUT(bloom_memtable_miss_count);
+  OUTPUT(bloom_sst_hit_count);
+  OUTPUT(bloom_sst_miss_count);
   return ss.str();
 #endif
 }