dhcpd capacity

[fadey]

Thanks for your reply
I've set 
ping-check off;
in the config file. Also now I'm running dhcpd with
dhcpd -d -f 2>/dev/null (I guess this way it will not write to disk on
every request). Yet what I see from a tcpdump trace is that the
situation has not improved:

0:42:31.941360 IP 192.168.0.2.67 > 10.2.0.1.67: BOOTP/DHCP, Reply,
length: 328
00:42:31.990506 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:0f:21:e7:50:bd, length: 436
00:42:32.055564 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1c:70, length: 512
00:42:32.311609 IP 10.2.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:16:17:aa:82:7a, length: 320
00:42:32.374133 IP 192.168.0.2.67 > 10.0.0.1.67: BOOTP/DHCP, Reply,
length: 300
00:42:32.672425 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1c:64, length: 512
00:42:32.779738 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1a:34, length: 512
00:42:33.004367 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1a:44, length: 524
00:42:33.108238 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1a:2c, length: 512
00:42:33.281421 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1b:e8, length: 524
00:42:33.621956 IP 192.168.0.2.67 > 10.0.0.1.67: BOOTP/DHCP, Reply,
length: 305
00:42:33.711499 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1a:1c, length: 512
00:42:33.904487 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:11:e6:f2:39:26, length: 424
00:42:34.010164 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:11:e6:f2:3c:cc, length: 424
00:42:34.037943 IP 192.168.0.2.67 > 10.0.0.1.67: BOOTP/DHCP, Reply,
length: 305
00:42:34.070474 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1c:68, length: 512
00:42:34.153420 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1c:44, length: 512
00:42:34.194339 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1a:48, length: 524
00:42:34.199977 IP 10.0.0.196.1055 > 192.168.0.2.69:  35 RRQ
"SA-EPX2203-VODA-BASICA.cfg" octet
00:42:34.418991 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:0f:21:e7:4f:db, length: 436
00:42:34.444522 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:0e:5c:a9:5e:a0, length: 523
00:42:34.481865 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:11:e6:f1:eb:6a, length: 436
00:42:34.649512 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:0f:21:e7:51:99, length: 436
00:42:34.696330 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:11:e6:f1:ea:ea, length: 424
00:42:34.761267 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:0e:5c:a9:5e:b2, length: 523
00:42:34.985876 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:11:e6:f1:eb:74, length: 424
00:42:35.099862 IP 192.168.0.2.67 > 10.2.0.1.67: BOOTP/DHCP, Reply,
length: 300
00:42:35.233276 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:0f:21:e7:4f:23, length: 436
00:42:35.595500 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1c:78, length: 512
00:42:35.702524 IP 10.2.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1a:22, length: 484
00:42:35.818158 IP 10.2.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:0e:5c:a6:68:7c, length: 300
00:42:35.991982 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:0f:21:e7:50:bd, length: 436
00:42:36.138845 IP 192.168.0.2.67 > 10.0.0.1.67: BOOTP/DHCP, Reply,
length: 300
00:42:36.281502 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1b:e8, length: 524
00:42:36.709991 IP 10.0.0.1.67 > 192.168.0.2.67: BOOTP/DHCP, Request
from 00:14:f8:8d:1c:5c, length: 512
00:42:36.776847 IP 192.168.0.2.67 > 10.2.0.1.67: BOOTP/DHCP, Reply,
length: 329


Just wondering if anyone has experienced this before. I've got just 200
cable modems in my network. I guess there are networks MUCH bigger then
that which are using isc dhcpd


> fadey wrote:
> 
> >I'm experiencing a problem with dhcpd in a cable network. Once cable
> >modems are reset (I have about 200), a storm of packets is flooding
> >dhcpd and it can't serve the networking  information to all of cable
> >modems at once. For about every 5 DISCOVERs I see 1 OFFER (with
> >tcpdump). So, I guess that either some OFFERSs are being dropped or by
> >the time they reach a cable modem a timeout opccurs.
> >I'm wondering if there is something in the configuration that can
> >influence the responce time of dhcpd.
> 
> Check the archives, there's been a number of threads about server 
> performance over the last few months. A few tips :
> 
> ping-before-offer : you can turn this off
> 
> logging : is syslog configured to write logs synchronously ? If so, 
> then this slows things down quite a bit (every log created results in 
> a disk write). Setting this to async helps quite a bit.
> 
> Also, some people have experimented with putting various things on 
> ram disk - but it may not be safe against reboots.
> 
> 
>