Hi,
I've just got my dsl line reprovisioned for Annex M compatibility and
the line speed showing currently on the modem is 20Mbps downstream with
2Mbps upstream.
Upon the reprovisioning I got ask to do a speed test by my ISP and only
~>5Mbps was shown for downstream while upstream was shown as ~700kbps.
The system in place is VDSL2+/ADSL2+(Annex M) router running in RFC2684
ATM to Ethernet bridge mode, linked to a Sun Microsystems Netra T105 -
440MHz, 360MB RAM box. With the Netra I have tested various routing
capacities including inter-vlan and dynamic running OSPF and it managed
between 80-90Mbps transfer rates for large files.
I'm using pppoe(4) with config pretty much taken from the man page:
inet 0.0.0.0 255.255.255.255 NONE mtu 1492 \
pppoedev hme0 authproto chap \
authname <authuser> authkey <authkey> \
up
dest 0.0.0.1
!/sbin/route add default -ifp pppoe0 0.0.0.1
running over the hme0 interface:
up mtu 1500
Everything is working fine and I have internet connectivity.
Packet Filter is of course taking care of the NAT/PAT side of things and
the firewalling.
Doing a bit of research on the net I came across a few postings:
http://forums.whirlpool.net.au/archive/481579
http://forums.whirlpool.net.au/forum-replies.cfm?t=116165
http://comments.gmane.org/gmane.os.openbsd.misc/196051
which suggested to edit the /etc/sysctl.conf file with some values that
aren't available in OBSD5.1 as they are now allocated dynamically.
This is what I've got towards the bottom of my sysctl.conf file:
net.inet.tcp.mssdflt=1452
#net.inet.tcp.recvspace=131072
#net.inet.tcp.sendspace=131072
net.inet.udp.recvspace=139264
net.inet.udp.sendspace=139264
net.inet.ip.mforwarding=1
ddb.panic=0
kern.bufcachepercent=60
net.inet.ip.ifq.maxlen=512
net.inet.tcp.rfc3390=1
I have also made the necessary adjustments to the pf.conf file:
match on $ext_if scrub (max-mss 1440)
however, my speeds are still really low :-(
The router/modem I'm using is a Cisco 887VA which should be pretty
high-performance but I don't trust Cisco's with NAT as they keep running
out of memory on me when the number of connections starts going up; then
they crash!
Is there anything else I can do to get better performance or this simply
it for the Netra?
I've tried checking with all kinds of tools like nload and darkstat for
bandwidth which confirms what the speed-test site claimed.
Running netstat -m I get:
# netstat -m
111 mbufs in use:
20 mbufs allocated to data
62 mbufs allocated to packet headers
29 mbufs allocated to socket names and addresses
19/88/4096 mbuf 2048 byte clusters in use (current/peak/max)
0/8/4096 mbuf 4096 byte clusters in use (current/peak/max)
0/8/4096 mbuf 8192 byte clusters in use (current/peak/max)
0/14/4102 mbuf 9216 byte clusters in use (current/peak/max)
0/8/4096 mbuf 12288 byte clusters in use (current/peak/max)
0/8/4096 mbuf 16384 byte clusters in use (current/peak/max)
0/8/4096 mbuf 65536 byte clusters in use (current/peak/max)
536 Kbytes allocated to network (12% in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines
vmstat -i shows:
# vmstat -i
interrupt total rate
com0 4988 1
hme0 82113 19
siop0 9026 2
hme1 86683 21
clock 412347 99
Total 595157 144
Netstat -i doesn't come up with any errors on any interfaces
either....... and finally top doesn't show much cpu usage at all either.
Surely the Netra must be able to do more then 4.5Mbps down with 0.7Mbps
up??? What about my LAN routing test which got 80-90Mbps?
Can anyone suggest anything?
Thanks.
