Telstra ADSL questions (pppoe technical)

[Eyal Lebedinsky]

Martijn van Oosterhout wrote:
> 
> I think it says somewhere that they consider 1M = 1000K.

To put it clearly, 1M=1,000,000?

> > 2) I see ADSL traffic on eth0 and on ppp0. I expected to see the raw
> >    volume on eth0, with the net on ppp0 but this is not what I see.
> >    While the number of packets on both is identical, the volume on
> >    ppp0 os higher. Can anyone explain?
> 
> The packets may be compressed, meaning that the traffic on eth0 will be less
> than on ppp0.

How can I tell? I am happy to turn on some debug for a while while
testing this.
 
> >    BTW, I xfered a large file (15,751,285) and eth0 had a rcv of
> >    16,644,696 while ppp0 shows 19,001,215. Reasonable? The xfer
> >    rcvd 11,358 packets and sent 5,758.
> 
> 20% overhead. Seems high. I would have thought more like 5-10%. Mind you,
> who knows what else went over the link at the time.

Nothing else of course. The whole point of my test was to do nothing
else
during the 5 minutes of the test. I positively know there was no other
activity.

> Well, there is no speed guarentee. And going by the 1K = 1000 byte rules,
> 51K is about right.

How? if one byte is 8 bits (well, call me crazy!) then 512,000 bits/sec
is 64,000 bytes/sec, right? 51000/64000= 20% missing.

Also, I am not saying I sometimes not get more that 51KBps, but I never
saw
anything better. In other words the line is limited to this speed, which
I
think is lower than the reasonable expectation from the plan.

> There are various types of compression. The actual bodies of packets can be
> compressed and the headers can be compressed. They are independant. On the
> link here it appears that there is VJ header compression but the bodies are
> uncompressed...

If I can turn anything on to check this then I will. my fw runs kernel
2.2.16
so I have:

# ls /proc/net 
dev_mcast      wireless       dev_stat       dev            bridge
psched         udp            tcp            sockstat       netstat
snmp           raw            ip_mr_cache    ip_mr_vif      ip_masq
ip_masquerade  ip_fwnames     ip_fwchains    igmp           rt_acct
rt_cache       route          arp            tr_rif         unix
netlink

Any of these of use?

Ta

--
Eyal Lebedinsky (eyal at eyal.emu.id.au) <http://samba.anu.edu.au/eyal/>