Skip to content

N
neo
Commit 090f4ac7 authored by Kirill Smelkov's avatar Kirill Smelkov
Browse files
Options

X Time after reducing RX latencies on teco.eth0 and deco.eth0 

200μs of NEO/go without sha1 case is still questionable:

T(transmit request) = 10μs (deco rx throttle) + t(link-tx-request)
T(transmit answer)  = 10μs·4.5 (~ 4.5 KB answer for 4K ZData) + teco.rx-delay (= 3μs ?)

	||
	vv

T(link) ~= 60μs

T(teco-local,no-sha1,NEO/go) was ~= 100μs (see time-sep17-teco.txt)
T(deco-local,no-sha1,NEO/go) was ~=  20μs (see time-sep17-deco.txt)

	||
	vv

T(teco-deco,no-sha1,NEO/go) = 100μs/2 + 20μs/2 = 60μs ?

if so RTT(getobject) should be 60+60 = 120μs  but we have 200μs.
parent 73b78d44
Hide whitespace changes
Inline Side-by-side
Showing with 115 additions and 0 deletions
go/time-sep19-teco4deco4.txt 0 → 100644
View file @ 090f4ac7
- teco & deco connected directly without switch with manually set IPv4 addresses
- teco: eth0: rx-usecs↓ to 3μs (was 20μs) (ethtool -C ...)
- deco: eth0: rx-usecs 3 (dynamic conservative; see e1000e docs) -> 10 (=10μs) (ethtool -C)
---- 8< ----
>>> bench-cluster neotest@teco4:qq
# server:
# Tue, 19 Sep 2017 20:14:06 +0300
# kirr@deco4 (192.168.0.2)
# Intel(R) Core(TM) i7-6600U CPU @ 2.60GHz
# Linux deco4 4.12.0-1-amd64 #1 SMP Debian 4.12.6-1 (2017-08-12) x86_64 GNU/Linux
# Python 2.7.13
# go version go1.9 linux/amd64
# mysqld Ver 10.1.26-MariaDB-1 for debian-linux-gnu on x86_64 (Debian unstable)
# neo : v1.8-1212-g73b78d44
# zodb : 5.2.0-9-gf0fe34791
# zeo : 4.2.0b1-336-gd5e82d48
# mysqlclient : 1.3.10
# wendelin.core : v0.11-4-g38fbc83
# client:
# Tue, 19 Sep 2017 20:14:09 +0300
# neotest@teco.navytux.spb.ru (2401:5180:0:25::1)
# Intel(R) Core(TM)2 CPU T7200 @ 2.00GHz
# Linux teco 4.12.0-1-amd64 #1 SMP Debian 4.12.6-1 (2017-08-12) x86_64 GNU/Linux
# Python 2.7.13
# go version go1.9 linux/amd64
# mysqld Ver 10.1.26-MariaDB-1 for debian-linux-gnu on x86_64 (Debian unstable)
2017-09-19 20:14:09 139822357619584 [Note] Using unique option prefix 'key_buffer' is error-prone and can break in the future. Please use the full name 'key_buffer_size' instead.
# neo : v1.3-1608-g60b9e4ba
# zodb : 5.3.0
# zeo : 5.1.0
# mysqlclient : 1.3.12
# wendelin.core : 0.11
# link latency:
# deco4 ⇄ teco4
PING teco4 (192.168.0.3) 56(84) bytes of data.
--- teco4 ping statistics ---
44751 packets transmitted, 44750 received, 0% packet loss, time 3000ms
rtt min/avg/max/mdev = 0.047/0.063/0.733/0.009 ms, ipg/ewma 0.067/0.064 ms
# teco4 ⇄ deco4
PING 192.168.0.2 (192.168.0.2) 56(84) bytes of data.
--- 192.168.0.2 ping statistics ---
35704 packets transmitted, 35703 received, 0% packet loss, time 2999ms
rtt min/avg/max/mdev = 0.066/0.068/0.345/0.012 ms, ipg/ewma 0.084/0.067 ms
*** ZEO
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=8.808s (1036.3μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=8.765s (1031.2μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=8.653s (1018.0μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=8.750s (1029.4μs / object) x=zsha1.py
(skipping zhash.go on ZEO -- Cgo does not support zeo:// protocol)
*** NEO/py sqlite
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=7.661s (901.3μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=7.611s (895.4μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=7.692s (904.9μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=7.632s (897.9μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=6.419480019s (755.232µs / object) x=zsha1.go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=6.388401347s (751.576µs / object) x=zsha1.go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=6.400416003s (752.99µs / object) x=zsha1.go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=6.432737641s (756.792µs / object) x=zsha1.go
2017-09-19 20:15:56.7475 ERROR NEO [ app: 91] primary master is down
Cluster state changed
*** NEO/py sql
2017-09-19 20:15:56 140451902690176 [Note] mysqld (mysqld 10.1.26-MariaDB-1) starting as process 11353 ...
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=10.151s (1194.3μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=9.221s (1084.8μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=9.692s (1140.2μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=9.255s (1088.9μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=8.809615594s (1.036425ms / object) x=zsha1.go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=9.130543556s (1.074181ms / object) x=zsha1.go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=9.410569668s (1.107125ms / object) x=zsha1.go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=8.156887549s (959.633µs / object) x=zsha1.go
2017-09-19 20:17:15.8135 ERROR NEO [ app: 91] primary master is down
Cluster state changed
*** NEO/go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=6.710s (789.4μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=6.708s (789.1μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=6.742s (793.2μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=6.674s (785.2μs / object) x=zsha1.py
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=2.134350325s (251.1µs / object) x=zsha1.go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=2.07626738s (244.266µs / object) x=zsha1.go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=2.055636684s (241.839µs / object) x=zsha1.go
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=2.025346659s (238.276µs / object) x=zsha1.go
2017-09-19 20:17:57.2394 ERROR NEO [ app: 91] primary master is down
Cluster state changed
E0919 20:17:57.248548 11398 storage.go:120] storage(192.168.0.2:5554): accept: accept tcp 192.168.0.2:5554: use of closed network connection
2017/09/19 20:17:57 talk master([192.168.0.2]:5552): context canceled
*** NEO/go (sha1 disabled)
# NEO/go/storage: skipping SHA1 computations
# NEO/go/client: skipping SHA1 checks
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=1.706528952s (200.768µs / object) x=zsha1.go
# NEO/go/client: skipping SHA1 checks
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=1.696704379s (199.612µs / object) x=zsha1.go
# NEO/go/client: skipping SHA1 checks
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=1.722048663s (202.593µs / object) x=zsha1.go
# NEO/go/client: skipping SHA1 checks
crc32:bf9deea9 ; oid=0..8499 nread=34134938 t=1.680703461s (197.729µs / object) x=zsha1.go
2017-09-19 20:18:05.4310 ERROR NEO [ app: 91] primary master is down
Cluster state changed
E0919 20:18:05.440081 11413 storage.go:120] storage(192.168.0.2:5554): accept: accept tcp 192.168.0.2:5554: use of closed network connection
2017/09/19 20:18:05 talk master([192.168.0.2]:5552): context canceled
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7