hi Dormando, launch params:
memcached -d -p 11211 -u nobody -m 1024 -c 128000 -P /var/run/memcached/memcached.pid -r -t 8 -v (ya, i know about 128K :) ) multiconnection performance is a bit better in .21. See below. I am still going to build .24 (i see big improvement there by release notes, but as i understand mostly for multi connection case due removing global lock) .13: [dsamoy...@dsamoylov.dev twemperf (master)]$ src/mcperf --linger=0 --call-rate=0 --num-calls=100 --conn-rate=0 --num-conns=1000 --sizes=d1 Total: connections 1000 requests 100000 responses 100000 test-duration 12.138 s Connection rate: 82.4 conn/s (12.1 ms/conn <= 1 concurrent connections) Connection time [ms]: avg 12.1 min 5.9 max 108.5 stddev 7.06 Connect time [ms]: avg 0.0 min 0.0 max 0.2 stddev 0.01 Request rate: 8238.5 req/s (0.1 ms/req) Request size [B]: avg 28.0 min 28.0 max 28.0 stddev 0.00 Response rate: 8238.5 rsp/s (0.1 ms/rsp) Response size [B]: avg 8.0 min 8.0 max 8.0 stddev 0.00 Response time [ms]: avg 0.1 min 0.0 max 23.8 stddev 0.00 Response time [ms]: p25 1.0 p50 1.0 p75 1.0 Response time [ms]: p95 1.0 p99 1.0 p999 5.0 Response type: stored 100000 not_stored 0 exists 0 not_found 0 Response type: num 0 deleted 0 end 0 value 0 Response type: error 0 client_error 0 server_error 0 Errors: total 0 client-timo 0 socket-timo 0 connrefused 0 connreset 0 Errors: fd-unavail 0 ftab-full 0 addrunavail 0 other 0 CPU time [s]: user 2.73 system 6.62 (user 22.5% system 54.5% total 77.0%) Net I/O: bytes 3.4 MB rate 289.6 KB/s (2.4*10^6 bps) .21 [dsamoy...@dsamoylov.dev twemperf (master)]$ src/mcperf --linger=0 --call-rate=0 --num-calls=100 --conn-rate=0 --num-conns=1000 --sizes=d1 Total: connections 1000 requests 100000 responses 100000 test-duration 11.458 s (11.722 s, 11.494 s) Connection rate: 87.3 conn/s (11.5 ms/conn <= 1 concurrent connections) Connection time [ms]: avg 11.5 min 5.8 max 65.8 stddev 5.32 Connect time [ms]: avg 0.0 min 0.0 max 0.3 stddev 0.01 Request rate: 8727.5 req/s (0.1 ms/req) Request size [B]: avg 28.0 min 28.0 max 28.0 stddev 0.00 Response rate: 8727.5 rsp/s (0.1 ms/rsp) Response size [B]: avg 8.0 min 8.0 max 8.0 stddev 0.00 Response time [ms]: avg 0.1 min 0.0 max 31.3 stddev 0.00 Response time [ms]: p25 1.0 p50 1.0 p75 1.0 Response time [ms]: p95 1.0 p99 1.0 p999 4.0 Response type: stored 100000 not_stored 0 exists 0 not_found 0 Response type: num 0 deleted 0 end 0 value 0 Response type: error 0 client_error 0 server_error 0 Errors: total 0 client-timo 0 socket-timo 0 connrefused 0 connreset 0 Errors: fd-unavail 0 ftab-full 0 addrunavail 0 other 0 CPU time [s]: user 2.78 system 6.54 (user 24.3% system 57.1% total 81.4%) Net I/O: bytes 3.4 MB rate 306.8 KB/s (2.5*10^6 bps) On Tuesday, July 7, 2015 at 9:40:29 PM UTC-7, Dormando wrote: > > What happens with more than one connection? A lot of things changed to > increase the scalability of it but single thread might be meh. > > What're the memcached start args? What exactly is that test doing? > gets/sets? Is .24 any better? > > On Tue, 7 Jul 2015, Denis Samoylov wrote: > > > hi, > > Does anybody know what can change between .13 ( STAT version 1.4.13, > STAT libevent 1.4.13-stable) and .21 (STAT version 1.4.21 STAT libevent > 2.0.21-stable)? > > > > memcached became about 15-20% slower. We discovered this in production > but there are too many moving parts. So I used twemperf ( > https://github.com/twitter/twemperf) to measure on dev server > > > > > > result for .13 (repeated many times) > > > > [dsamoylov.dev twemperf (master)]$ src/mcperf --linger=0 --call-rate=0 > --num-calls=100000 --conn-rate=0 --num-conns=1 --sizes=d1 > > > > Total: connections 1 requests 100000 responses 100000 test-duration > 9.731 s (10.160 s, 9.981 s) > > > > Connection rate: 0.1 conn/s (9731.1 ms/conn <= 1 concurrent connections) > > Connection time [ms]: avg 9731.1 min 9731.1 max 9731.1 stddev 0.00 > > Connect time [ms]: avg 0.2 min 0.2 max 0.2 stddev 0.00 > > > > Request rate: 10276.3 req/s (0.1 ms/req) > > Request size [B]: avg 28.0 min 28.0 max 28.0 stddev 0.00 > > > > Response rate: 10276.3 rsp/s (0.1 ms/rsp) > > Response size [B]: avg 8.0 min 8.0 max 8.0 stddev 0.00 > > Response time [ms]: avg 0.1 min 0.0 max 18.0 stddev 0.00 > > Response time [ms]: p25 1.0 p50 1.0 p75 1.0 > > Response time [ms]: p95 1.0 p99 1.0 p999 4.0 > > Response type: stored 100000 not_stored 0 exists 0 not_found 0 > > Response type: num 0 deleted 0 end 0 value 0 > > Response type: error 0 client_error 0 server_error 0 > > > > Errors: total 0 client-timo 0 socket-timo 0 connrefused 0 connreset 0 > > Errors: fd-unavail 0 ftab-full 0 addrunavail 0 other 0 > > > > CPU time [s]: user 1.18 system 3.50 (user 12.2% system 36.0% total > 48.2%) > > Net I/O: bytes 3.4 MB rate 361.3 KB/s (3.0*10^6 bps) > > > > > > result for .21 (repeated many times) > > [dsamoylov.dev twemperf (master)]$ src/mcperf --linger=0 --call-rate=0 > --num-calls=100000 --conn-rate=0 --num-conns=1 --sizes=d1 > > > > Total: connections 1 requests 100000 responses 100000 test-duration > 12.328 s (11.230 s, 12.713 s) > > > > Connection rate: 0.1 conn/s (12328.4 ms/conn <= 1 concurrent > connections) > > Connection time [ms]: avg 12328.4 min 12328.4 max 12328.4 stddev 0.00 > > Connect time [ms]: avg 0.3 min 0.3 max 0.3 stddev 0.00 > > > > Request rate: 8111.4 req/s (0.1 ms/req) > > Request size [B]: avg 28.0 min 28.0 max 28.0 stddev 0.00 > > > > Response rate: 8111.4 rsp/s (0.1 ms/rsp) > > Response size [B]: avg 8.0 min 8.0 max 8.0 stddev 0.00 > > Response time [ms]: avg 0.1 min 0.0 max 28.3 stddev 0.00 > > Response time [ms]: p25 1.0 p50 1.0 p75 1.0 > > Response time [ms]: p95 1.0 p99 1.0 p999 5.0 > > Response type: stored 100000 not_stored 0 exists 0 not_found 0 > > Response type: num 0 deleted 0 end 0 value 0 > > Response type: error 0 client_error 0 server_error 0 > > > > Errors: total 0 client-timo 0 socket-timo 0 connrefused 0 connreset 0 > > Errors: fd-unavail 0 ftab-full 0 addrunavail 0 other 0 > > > > CPU time [s]: user 3.38 system 7.44 (user 27.4% system 60.3% total > 87.7%) > > Net I/O: bytes 3.4 MB rate 285.2 KB/s (2.3*10^6 bps) > > > > same server is used, clean just restarted memcached daemon. > > > > -- > > > > --- > > You received this message because you are subscribed to the Google > Groups "memcached" group. > > To unsubscribe from this group and stop receiving emails from it, send > an email to memcached+...@googlegroups.com <javascript:>. > > For more options, visit https://groups.google.com/d/optout. > > > > -- --- You received this message because you are subscribed to the Google Groups "memcached" group. To unsubscribe from this group and stop receiving emails from it, send an email to memcached+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
