Perf is a program that contains several micro-benchmarks related to RAMCloud. It includes all of the server code but runs as a stand-alone program, not in a cluster configuration. This page contains the output of that program at various points in time, so it provides a history of RAMCloud micro-benchmark performance.
Note: for consistency, please take all measurements on rc20 and be sure to compile with DEBUG=no.
atomicIntCmpX 6.23ns Atomic<int>::compareExchange atomicIntInc 5.85ns Atomic<int>::inc atomicIntInc 5.84ns std::atomic<int>::fetch_add atomicIntLoad 0.56ns Atomic<int>::load atomicIntStore 0.56ns Atomic<int>::store atomicIntXchg 5.84ns Atomic<int>::exchange bMutexNoBlock 14.46ns std::mutex lock/unlock (no blocking) bufferAppendCopy1 8.09ns appendCopy 1 byte to a buffer bufferAppendCopy50 8.96ns appendCopy 50 bytes to a buffer bufferAppendCopy100 9.09ns appendCopy 100 bytes to a buffer bufferAppendCopy250 18.08ns appendCopy 250 bytes to a buffer bufferAppendCopy500 22.56ns appendCopy 500 bytes to a buffer bufferAppendExternal1 8.24ns appendExternal 1 byte to a buffer bufferAppendExternal50 7.87ns appendExternal 50 bytes to a buffer bufferAppendExternal100 7.95ns appendExternal 100 bytes to a buffer bufferAppendExternal250 7.79ns appendExternal 250 bytes to a buffer bufferAppendExternal500 7.72ns appendExternal 500 bytes to a buffer bufferBasic 11.13ns buffer create, add one chunk, delete bufferBasicAlloc 11.99ns buffer create, alloc block in chunk, delete bufferBasicCopy 12.00ns buffer create, copy small block, delete bufferCopy 14.20ns copy out 2 small chunks from buffer bufferExtendChunk 4.56ns buffer add onto existing chunk bufferGetStart 3.34ns Buffer::getStart bufferConstruct 5.84ns buffer stack allocation bufferReset 4.72ns Buffer::reset bufferCopyIterator2 6.67ns buffer iterate over 2 copied chunks, accessing 1 byte each bufferCopyIterator5 9.18ns buffer iterate over 5 copied chunks, accessing 1 byte each bufferExternalIterator2 9.06ns buffer iterate over 2 external chunks, accessing 1 byte each bufferExternalIterator5 23.24ns buffer iterate over 5 external chunks, accessing 1 byte each condPingPong 2.98us std::condition_variable round-trip cppAtomicExchg 5.83ns Exchange method on a C++ atomic_int cppAtomicLoad 0.65ns Read a C++ atomic_int cyclesToSeconds 6.13ns Convert a rdtsc result to (double) seconds cyclesToNanos 6.35ns Convert a rdtsc result to (uint64_t) nanoseconds dispatchPoll 9.21ns Dispatch::poll (no timers or pollers) div32 4.72ns 32-bit integer division instruction div64 20.15ns 64-bit integer division instruction functionCall 1.67ns Call a function that has not been inlined generateRandomNumber 16.56ns Call to randomNumberGenerator(x) genRandomString 1.49us Generate a random 100-byte value getThreadId 1.95ns Retrieve thread id via ThreadId::get getThreadIdSyscall 41.15ns Retrieve kernel thread id using syscall hashTableLookup 124.09ns Key lookup in a 1GB HashTable hashTableLookupPf 73.98ns Key lookup in a 1GB HashTable with prefetching lfence 2.92ns Lfence instruction lockInDispThrd 2.79ns Acquire/release Dispatch::Lock (in dispatch thread) lockNonDispThrd 279.30ns Acquire/release Dispatch::Lock (non-dispatch thread) mapCreate 22.63ns Create+delete entry in std::map mapLookup 16.02ns Lookup in std::map<uint64_t, uint64_t> memcpyCached100 14.55ns memcpy 100 bytes with hot/fixed dst and src memcpyCached1000 21.82ns memcpy 1000 bytes with hot/fixed dst and src memcpyCached10000 185.16ns memcpy 10000 bytes with hot/fixed dst and src memcpyCachedDst100 93.82ns memcpy 100 bytes with hot/fixed dst and cold src memcpyCachedDst1000 122.30ns memcpy 1000 bytes with hot/fixed dst and cold src memcpyCachedDst10000 355.23ns memcpy 10000 bytes with hot/fixed dst and cold src memcpyCold100 372.20ns memcpy 100 bytes with cold dst and src memcpyCold1000 655.00ns memcpy 1000 bytes with cold dst and src memcpyCold10000 3.18us memcpy 10000 bytes with cold dst and src murmur3 10.28ns 128-bit MurmurHash3 (64-bit optimised) on 1 byte of data murmur3 49.72ns 128-bit MurmurHash3 hash (64-bit optimised) on 256 bytes of data objectPoolAlloc 29.16ns Cost of new allocations from an ObjectPool (no destroys) objectPoolRealloc 4.48ns Cost of ObjectPool allocation after destroying an object pingConditionVar 1.56us Round-trip ping with std::condition_variable prefetch 35.20ns Prefetch instruction rdtsc 6.67ns Read the fine-grain cycle counter segmentEntrySort 8.14ms Sort a Segment full of avg. 100-byte Objects by age segmentIterator 1.97ms Iterate a Segment full of avg. 100-byte Objects sessionRefCount 12.55ns Create/delete SessionRef serialize 55.34ns cpuid instruction for serialize sfence 1.39ns Sfence instruction spinLock 18.90ns Acquire/release SpinLock startStopTimer 56.46ns Start and stop a Dispatch::Timer spawnThread 8.90us Start and stop a thread throwInt 2.02us Throw an int throwIntNL 2.48us Throw an int in a function call throwException 1.97us Throw an Exception throwExceptionNL 2.78us Throw an Exception in a function call throwSwitch 6.05us Throw an Exception using ClientException::throwException timeTrace 15.93ns Record an event using TimeTrace unorderedMapCreate 72.63ns Create+delete entry in unordered_map unorderedMapLookup 13.07ns Lookup in std::unordered_map<uint64_t, uint64_t> vectorPushPop 3.80ns Push and pop a std::vector |
atomicIntCmpX 6.21ns Atomic<int>::compareExchange atomicIntInc 5.98ns Atomic<int>::inc atomicIntLoad 0.56ns Atomic<int>::load atomicIntStore 0.56ns Atomic<int>::store atomicIntXchg 5.97ns Atomic<int>::exchange bMutexNoBlock 14.77ns std::mutex lock/unlock (no blocking) bufferBasic 11.42ns buffer create, add one chunk, delete bufferBasicAlloc 14.62ns buffer create, alloc block in chunk, delete bufferBasicCopy 13.24ns buffer create, copy small block, delete bufferCopy 14.76ns copy out 2 small chunks from buffer bufferExtendChunk 5.69ns buffer add onto existing chunk bufferGetStart 3.34ns Buffer::getStart bufferIterator 22.57ns iterate over buffer with 5 chunks condPingPong 4.04us std::condition_variable round-trip cppAtomicExchg 5.87ns Exchange method on a C++ atomic_int cppAtomicLoad 13.50ns Read a C++ atomic_int cyclesToSeconds 6.97ns Convert a rdtsc result to (double) seconds cyclesToNanos 9.75ns Convert a rdtsc result to (uint64_t) nanoseconds dispatchPoll 8.63ns Dispatch::poll (no timers or pollers) div32 4.86ns 32-bit integer division instruction div64 20.30ns 64-bit integer division instruction functionCall 1.95ns Call a function that has not been inlined getThreadId 2.24ns Retrieve thread id via ThreadId::get hashTableLookup 128.86ns Key lookup in a 1GB HashTable hashTableLookupPf 78.07ns Key lookup in a 1GB HashTable with prefetching lfence 2.82ns Lfence instruction lockInDispThrd 4.06ns Acquire/release Dispatch::Lock (in dispatch thread) lockNonDispThrd 214.54ns Acquire/release Dispatch::Lock (non-dispatch thread) memcpy100 8.78ns Copy 100 bytes with memcpy memcpy1000 37.37ns Copy 1000 bytes with memcpy memcpy10000 191.07ns Copy 10000 bytes with memcpy murmur3 10.31ns 128-bit MurmurHash3 (64-bit optimised) on 1 byte of data murmur3 40.47ns 128-bit MurmurHash3 hash (64-bit optimised) on 256 bytes of data objectPoolAlloc 27.86ns Cost of new allocations from an ObjectPool (no destroys) objectPoolRealloc 4.65ns Cost of ObjectPool allocation after destroying an object prefetch 34.72ns Prefetch instruction rdtsc 6.70ns Read the fine-grain cycle counter segmentEntrySort 7.02ms Sort a Segment full of avg. 100-byte Objects by age segmentIterator 1.83ms Iterate a Segment full of avg. 100-byte Objects sfence 1.39ns Sfence instruction spinLock 16.17ns Acquire/release SpinLock startStopTimer 14.76ns Start and stop a Dispatch::Timer spawnThread 9.89us Start and stop a thread throwInt 1.91us Throw an int throwIntNL 2.60us Throw an int in a function call throwException 1.84us Throw an Exception throwExceptionNL 2.58us Throw an Exception in a function call throwSwitch 5.46us Throw an Exception using ClientException::throwException vectorPushPop 3.72ns Push and pop a std::vector |
atomicIntCmpX 6.29ns Atomic<int>::compareExchange atomicIntInc 5.93ns Atomic<int>::inc atomicIntLoad 0.56ns Atomic<int>::load atomicIntStore 0.59ns Atomic<int>::store atomicIntXchg 5.93ns Atomic<int>::exchange bMutexNoBlock 14.92ns std::mutex lock/unlock (no blocking) cppAtomicExchg 5.93ns Exchange method on a C++ atomic_int cppAtomicLoad 13.76ns Read a C++ atomic_int cyclesToSeconds 7.04ns Convert a rdtsc result to (double) seconds cyclesToNanos 9.86ns Convert a rdtsc result to (uint64_t) nanoseconds dispatchPoll 8.17ns Dispatch::poll (no timers or pollers) functionCall 1.98ns Call a function that has not been inlined getThreadId 1.98ns Retrieve thread id via ThreadId::get hashTableLookup 168.76ns Key lookup in a 1GB HashTable hashTableLookupPf 143.19ns Key lookup in a 1GB HashTable with prefetching lfence 2.90ns Lfence instruction lockInDispThrd 4.15ns Acquire/release Dispatch::Lock (in dispatch thread) lockNonDispThrd 4.05ns Acquire/release Dispatch::Lock (non-dispatch thread) murmur3 11.29ns 128-bit MurmurHash3 (64-bit optimised) on 1 byte of data murmur3 41.17ns 128-bit MurmurHash3 hash (64-bit optimised) on 256 bytes of data objectPoolAlloc 192.56ns Cost of new allocations from an ObjectPool (no destroys) objectPoolRealloc 4.42ns Cost of ObjectPool allocation after destroying an object prefetch 29.79ns Prefetch instruction rdtsc 6.80ns Read the fine-grain cycle counter segmentEntrySort 5.18ms Sort a Segment full of avg. 100-byte Objects by age segmentIterator 1.08ms Iterate a Segment full of avg. 100-byte Objects sfence 1.47ns Sfence instruction spinLock 16.34ns Acquire/release SpinLock startStopTimer 15.03ns Start and stop a Dispatch::Timer throwInt 1.97us Throw an int throwIntNL 2.44us Throw an int in a function call throwException 1.88us Throw an Exception throwExceptionNL 2.60us Throw an Exception in a function call throwSwitch 5.09us Throw an Exception using ClientException::throwException vectorPushPop 3.86ns Push and pop a std::vector |
atomicIntCmpX 6.26ns AtomicInt::compareExchange atomicIntInc 6.05ns AtomicInt::inc atomicIntLoad 0.57ns AtomicInt::load atomicIntStore 0.58ns AtomicInt::store atomicIntXchg 6.06ns AtomicInt::exchange bMutexNoBlock 14.66ns Boost mutex lock/unlock (no blocking) cppAtomicExchg 5.93ns Exchange method on a C++ atomic_int cppAtomicLoad 13.66ns Read a C++ atomic_int dispatchPoll 7.62ns Dispatch::poll (no timers or pollers) getThreadId 2.27ns Retrieve thread id via ThreadId::get lfence 2.84ns Lfence instruction lockInDispThrd 7.05ns Acquire/release Dispatch::Lock (in dispatch thread) lockNonDispThrd 252.79ns Acquire/release Dispatch::Lock (non-dispatch thread) sfence 1.50ns Sfence instruction spinLock 16.46ns Acquire/release SpinLock |
atomicIntCmpX 10.63ns AtomicInt::compareExchange atomicIntInc 8.71ns AtomicInt::inc atomicIntLoad 0.44ns AtomicInt::load atomicIntStore 0.43ns AtomicInt::store atomicIntXchg 7.83ns AtomicInt::exchange bMutexNoBlock 42.26ns Boost mutex lock/unlock (no blocking) cppAtomicExchg 7.97ns Exchange method on a C++ atomic_int cppAtomicLoad 7.69ns Read a C++ atomic_int cyclesToSeconds 10.42ns Convert a rdtsc result to (double) seconds cyclesToNanos 14.49ns Convert a rdtsc result to (uint64_t) nanoseconds dispatchPoll 20.41ns Dispatch::poll (no timers or pollers) functionCall 3.05ns Call a function that has not been inlined getThreadId 3.04ns Retrieve thread id via ThreadId::get lfence 4.79ns Lfence instruction lockInDispThrd 6.10ns Acquire/release Dispatch::Lock (in dispatch thread) lockNonDispThrd 283.43ns Acquire/release Dispatch::Lock (non-dispatch thread) sfence 3.60ns Sfence instruction spinLock 25.17ns Acquire/release SpinLock throwInt 3.32us Throw an int throwIntNL 4.54us Throw an int in a function call throwException 3.43us Throw an Exception throwExceptionNL 4.69us Throw an Exception in a function call throwSwitch 6.45us Throw an Exception using ClientException::throwException |