[RP-PPPoE] 1000 pppoe connections terminate at once cause high cpu load

[守得云开]

I have a pppoe-server running at a i3-3240 cpu and 4GB memory hp server, with no ip-up or ip-down scripts for test.

i have done 2 load tests:
A. 1000 connections dail up at once
B. 1000 connections terminate at once

test for A:
it's ok, uptime shows sysload is less than 1

test for B:
uptime command shows syslaod is up to 500 or more, that's very high. it's a bug or not?　is there any way to reduce sysload?


infos:
---------------------------------------------------------------------------------------------------------------------------------------
PPPoE-Server Version 3.11, Copyright (C) 2001-2009 Roaring Penguin Software Inc.

uname -a
Linux AxinChan 3.4.105 #2 SMP Tue Mar 3 16:56:17 CST 2015 x86_64 GNU/Linux

/usr/sbin/pppoe-server -k -O /etc/ppp/pppoe-server-options -I eth3 -C My_Bras -S My_Bras -L 10.8.0.1 -p /etc/ppp/pppoe-server.ipppool -X /var/run/bras.eth3.pid

cat /etc/ppp/pppoe-server-options 
auth
#persist
#debug
#kdebug 7
-pap
+chap
default-asyncmap
lcp-echo-interval 30
lcp-echo-failure 4
noipdefault
noipx
nodefaultroute
noproxyarp
noktune
netmask 255.255.255.255
nobsdcomp
nodeflate
nopcomp
nomppe
ipcp-accept-local
ipcp-accept-remote
logfile /var/log/bras.log
nolog
mtu 1492
mru 1492
#lcp-echo-adaptive
ms-dns 10.8.0.1
#plugin /etc/ppp/plugins/rp-pppoe.so
plugin /lib/pppd/2.4.7/radius.so
plugin /lib/pppd/2.4.7/radattr.so
radius-config-file /etc/ppp/radiusclient/radiusclient.conf



axinchan from china