#!/bin/sh . ../.function DOC_ROOT=benchmark/FrameworkBenchmarks/ULib/www rm -f benchmark/FrameworkBenchmarks/benchmark.log* \ /tmp/*.memusage.* /tmp/request* /tmp/response* \ err/FrameworkBenchmarks.err out/userver_*.out err/userver_*.err \ trace.*userver_*.[0-9]* object.*userver_*.[0-9]* stack.*userver_*.[0-9]* mempool.*userver_*.[0-9]* \ $DOC_ROOT/trace.*userver_*.[0-9]* $DOC_ROOT/object.*userver_*.[0-9]* $DOC_ROOT/stack.*userver_*.[0-9]* $DOC_ROOT/mempool.*userver_*.[0-9]* #UTRACE="0 100M 0" #UOBJDUMP="0 10M 100" #USIMERR="error.sim" #VALGRIND="valgrind -v --trace-children=yes" UMEMUSAGE=yes export UTRACE UOBJDUMP USIMERR VALGRIND UMEMUSAGE unset ORM_DRIVER ORM_OPTION export ORM_DRIVER ORM_OPTION UMEMPOOL # ---------------------------------------------------------------------------------------------------------------------------------------------------------- # PLAINTEXT # ---------------------------------------------------------------------------------------------------------------------------------------------------------- #UMEMPOOL="982,0,0,36,9846,-24,-23,1727,1151" #sed -i "s|TCP_LINGER_SET .*|TCP_LINGER_SET 0|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg #sed -i "s|LISTEN_BACKLOG .*|LISTEN_BACKLOG 16384|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg #sed -i "s|CLIENT_THRESHOLD .*|CLIENT_THRESHOLD 4000|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg #sed -i "s|CLIENT_FOR_PARALLELIZATION .*|CLIENT_FOR_PARALLELIZATION 8000|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg # ---------------------------------------------------------------------------------------------------------------------------------------------------------- # Running 15s test @ http://laptop:8080/plaintext # 4 threads and 16384 connections # Thread Stats Avg Stdev Max +/- Stdev # Latency 13.30s 2.61s 14.43s 87.33% # Req/Sec 3.06k 0.94k 5.07k 62.22% # Latency Distribution # 50% 14.38s # 75% 14.43s # 90% 14.43s # 99% 14.43s # 174245 requests in 15.00s, 23.93MB read # Socket errors: connect 0, read 143, write 0, timeout 5565 # Requests/sec: 11613.20 # Transfer/sec: 1.60MB # ---------------------------------------------------------------------------------------------------------------------------------------------------------- # JSON # ---------------------------------------------------------------------------------------------------------------------------------------------------------- UMEMPOOL="56,0,0,40,150,-24,-13,-20,0" sed -i "s|TCP_LINGER_SET .*|TCP_LINGER_SET 0|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg sed -i "s|LISTEN_BACKLOG .*|LISTEN_BACKLOG 256|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg #sed -i "s|CLIENT_THRESHOLD .*|CLIENT_THRESHOLD 50|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg #sed -i "s|CLIENT_FOR_PARALLELIZATION .*|CLIENT_FOR_PARALLELIZATION 100|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg # ---------------------------------------------------------------------------------------------------------------------------------------------------------- # Running 15s test @ http://laptop:8080/json # 4 threads and 256 connections # Thread Stats Avg Stdev Max +/- Stdev # Latency 18.26ms 3.84ms 234.44ms 91.85% # Req/Sec 3.48k 490.60 4.95k 67.32% # Latency Distribution # 50% 18.56ms # 75% 19.81ms # 90% 21.00ms # 99% 23.05ms # 205824 requests in 15.01s, 32.19MB read # Requests/sec: 13716.68 # Transfer/sec: 2.15MB # ---------------------------------------------------------------------------------------------------------------------------------------------------------- # DB # ---------------------------------------------------------------------------------------------------------------------------------------------------------- #ORM_DRIVER="sqlite" #ORM_OPTION="host=localhost dbname=../db/hello_world" #ORM_DRIVER="mysql" #ORM_OPTION="host=localhost user=benchmarkdbuser password=benchmarkdbpass character-set=utf8 dbname=hello_world" #ORM_DRIVER="pgsql" #ORM_OPTION="host=localhost user=benchmarkdbuser password=benchmarkdbpass dbname=hello_world" #UMEMPOOL="146,0,0,90,150,-22,-12,-20,0" #sed -i "s|TCP_LINGER_SET .*|TCP_LINGER_SET -2|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg #sed -i "s|LISTEN_BACKLOG .*|LISTEN_BACKLOG 256|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg #sed -i "s|CLIENT_THRESHOLD .*|CLIENT_THRESHOLD 80|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg #sed -i "s|CLIENT_FOR_PARALLELIZATION .*|CLIENT_FOR_PARALLELIZATION 100|g" benchmark/FrameworkBenchmarks/fbenchmark.cfg # ---------------------------------------------------------------------------------------------------------------------------------------------------------- # Running 15s test @ http://laptop:8080/fortune # 4 threads and 256 connections # Thread Stats Avg Stdev Max +/- Stdev # Latency 53.07ms 28.32ms 155.09ms 65.25% # Req/Sec 1.22k 164.62 1.57k 72.42% # Latency Distribution # 50% 51.15ms # 75% 72.30ms # 90% 88.69ms # 99% 134.90ms # 72743 requests in 15.00s, 94.28MB read # Requests/sec: 4849.05 # Transfer/sec: 6.28MB # # Running 15s test @ http://laptop:8080/db # 4 threads and 256 connections # Thread Stats Avg Stdev Max +/- Stdev # Latency 18.44ms 3.19ms 29.77ms 87.44% # Req/Sec 3.45k 439.79 4.98k 78.05% # Latency Distribution # 50% 18.80ms # 75% 20.13ms # 90% 21.27ms # 99% 23.43ms # 203711 requests in 15.00s, 32.60MB read # Requests/sec: 13583.13 # Transfer/sec: 2.17MB # # Running 15s test @ http://laptop:8080/query?queries=20 # 4 threads and 256 connections # Thread Stats Avg Stdev Max +/- Stdev # Latency 31.59ms 5.02ms 52.48ms 68.50% # Req/Sec 2.03k 167.87 2.45k 64.85% # Latency Distribution # 50% 31.69ms # 75% 34.52ms # 90% 38.22ms # 99% 43.36ms # 121069 requests in 15.00s, 89.44MB read # Requests/sec: 8072.57 # Transfer/sec: 5.96MB # ---------------------------------------------------------------------------------------------------------------------------------------------------------- DIR_CMD="../../examples/userver" prepare_usp mkdir -p $DOC_ROOT # ln -sf ../../../docroot/ws; \ # ln -sf ../../../docroot/servlet; \ # ln -sf ../../../docroot/100.html; \ # ln -sf ../../../docroot/1000.html; \ if [ "$TERM" != "cygwin" ]; then ( cd $DOC_ROOT; rm -f *; \ ln -sf ../../../../../../src/ulib/net/server/plugin/usp/.libs/json.so; \ ln -sf ../../../../../../src/ulib/net/server/plugin/usp/.libs/plaintext.so; \ ln -sf ../../../../../../src/ulib/net/server/plugin/usp/.libs/db.so; \ ln -sf ../../../../../../src/ulib/net/server/plugin/usp/.libs/update.so; \ ln -sf ../../../../../../src/ulib/net/server/plugin/usp/.libs/query.so; \ ln -sf ../../../../../../src/ulib/net/server/plugin/usp/.libs/fortune.so ) fi # ---------------------------------------------------------------------------------------------------------------------------------------------------------- #ulimit -n 100000 #echo 0 > /proc/sys/net/ipv4/tcp_syncookies # ---------------------------------------------------------------------------------------------------------------------------------------------------------- # cat /proc/sys/net/ipv4/tcp_mem # ------------------------------ # 47556 63410 95112 # ------------------------------------------- # low threshold => 47556 * 4096 = 185M # memory pressure => 63410 * 4096 = 247M # Out of socket memory => 95112 * 4096 = 371M # ------------------------------------------- # cat /proc/net/sockstat # cat /proc/sys/net/ipv4/tcp_max_orphans => 131072 # ---------------------------------------------------------------------------------------------------------------------------------------------------------- # So in this case we have 21564 orphans. That doesn't seem very close to 65536... Yet, if you look once more at the code above that prints the warning, # you'll see that there is this shift variable that has a value between 0 and 2, and that the check is testing if # (orphans << shift > sysctl_tcp_max_orphans). What this means is that in certain cases, the kernel decides to penalize some sockets more, and it does # so by multiplying the number of orphans by 2x or 4x to artificially increase the "score" of the "bad socket" to penalize. The problem is that due to # the way this is implemented, you can see a worrisome "Out of socket memory" error when in fact you're still 4x below the limit and you just had a # couple "bad sockets" (which happens frequently when you have an Internet facing service). So unfortunately that means that you need to tune up the # maximum number of orphan sockets even if you're 2x or 4x away from the threshold. What value is reasonable for you depends on your situation at hand. # Observe how the count of orphans in /proc/net/sockstat is changing when your server is at peak traffic, multiply that value by 4, round it up a bit # to have a nice value, and set it. You can set it by doing a echo of the new value in /proc/sys/net/ipv4/tcp_max_orphans, and don't forget to update # the value of net.ipv4.tcp_max_orphans in /etc/sysctl.conf so that your change persists across reboots. # ---------------------------------------------------------------------------------------------------------------------------------------------------------- # A server that uses SYN cookies doesn't have to drop connections when its SYN queue fills up. # Instead it sends back a SYN+ACK, exactly as if the SYN queue had been larger. # (Exceptions: the server must reject TCP options such as large windows, and it must use one of the # eight MSS values that it can encode.) When the server receives an ACK, it checks that the secret # function works for a recent value of t, and then rebuilds the SYN queue entry from the encoded MSS. # ---------------------------------------------------------------------------------------------------------------------------------------------------------- #echo 1024 > /proc/sys/net/core/somaxconn #echo 1 > /proc/sys/net/ipv4/tcp_syncookies #echo 524288 > /proc/sys/net/ipv4/tcp_max_orphans #echo 2 > /proc/sys/net/ipv4/tcp_synack_retries # 5 -> 2 == 21 sec (Total time to keep half-open connections in the backlog queue) #echo 0 > /proc/sys/kernel/printk_ratelimit # 5 #echo 0 > /proc/sys/kernel/printk_ratelimit_burst # 10 # -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- #STRACE=$TRUSS #VALGRIND="valgrind --tool=exp-dhat" #MUDFLAP_OPTIONS="-ignore-reads -backtrace=8" echo 16384 > /proc/sys/net/core/somaxconn start_prg_background userver_tcp -c benchmark/FrameworkBenchmarks/fbenchmark.cfg #run command on another computer #ab -n 100000 -c10 http://stefano/servlet/benchmarking?name=stefano #ab -n 100000 -c10 https://stefano/servlet/benchmarking?name=stefano #$SLEEP #killall userver_tcp userver_ssl mv err/userver_tcp.err err/FrameworkBenchmarks.err #mv err/userver_ssl.err err/benchmarking.err echo "PID = `cat benchmark/FrameworkBenchmarks/ULib/userver_tcp.pid`" # grep -v 'EAGAIN\|EPIPE\|ENOTCONN\|ECONNRESET' err/benchmarking.err # gprof -b ../../examples/userver/userver_tcp gmon.out >profile.out 2>/dev/null