Performance on udp socket

garcol at postino.it garcol at postino.it
Fri Jan 19 11:38:12 UTC 2007


Hi, 
    I've seen a packets drop with high rate of radius authentication request
(UDP 150 byte)
The software is radiator (perl), and it seems not to be dependent on CPU speed.
How is possible to avoid this with a more accurate tuning?


#netstat -s -p udp

udp:
1917 datagrams received
0 with incomplete header
0 with bad data length field
0 with bad checksum
0 with no checksum
0 dropped due to no socket
0 broadcast/multicast datagrams dropped due to no socket
557 dropped due to full socket buffers
0 not for hashed pcb
1360 delivered
1226 datagrams output


#netstat -na

Active Internet connections (including servers)
Proto Recv-Q Send-Q  Local Address          Foreign Address        (state)
udp4    5162      0  *.1813                 *.*                    
udp4   13615      0  *.1812                 *.*                    


#sysctl -a

...
udpcb:           192,     1992,      8,     34,    26713
...
net.inet.udp.maxdgram: 9216
net.inet.udp.recvspace: 42080
...
kern.ipc.maxsockbuf: 1048576
kern.ipc.nmbclusters: 1488


#netstat -s -p udp

udp:
4333 datagrams received
0 with incomplete header
0 with bad data length field
0 with bad checksum
0 with no checksum
0 dropped due to no socket
0 broadcast/multicast datagrams dropped due to no socket
724 dropped due to full socket buffers
0 not for hashed pcb
3609 delivered
3582 datagrams output


udp:
6366 datagrams received
0 with incomplete header
0 with bad data length field
0 with bad checksum
1 with no checksum
4 dropped due to no socket
8 broadcast/multicast datagrams dropped due to no socket
724 dropped due to full socket buffers
0 not for hashed pcb
6293 delivered
6315 datagrams output



netstat -m
158/720/5952 mbufs in use (current/peak/max):
        135 mbufs allocated to data
        23 mbufs allocated to packet headers
128/368/1488 mbuf clusters in use (current/peak/max)
916 Kbytes allocated to network (20% of mb_map in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines




More information about the freebsd-performance mailing list