John Linton
............or will BitTorrent commit commercial suicide by ending copyright theft in one fell swoop by doing what AFACT and all the other copyright owner's agencies and all National Governments have failed to do by making all P2P piracy the number one criminal act on all ISPs/Carriers prohibit lists? It will be interesting to see if the Australian ISPs now reverse their attitude to forwarding copyright infringement notices and, Heaven forbid, even start discontinuing P2P pirates who will be destroying their networks - or at least destroying their cost models for traffic.
......Another Night - Another Problem From Left Field
Managing a small business means you work for almost as many hours as you are awake on most days and so I was alerted to this article by my lap top 'beeping' with an email from one of Exetel's volunteer forum admins as I was finishing off a pleasant bottle of red and half watching something or other on Fox:
http://www.theregister.co.uk/2008/12/01/richard_bennett_utorrent_udp/
I read the article and immediately sent it to Steve who replied a couple of hours later confirming the article writer's view that if BitTorrent did switch to using UDP instead of TCP then there would be serious ramifications including bringing legitimate use of the internet to a halt. His reply to me, in part, was:
"This is my initial analysis:
1. P2P will no longer be controllable via the existing methods of
bandwidth throttling, because UDP does not have the control bits that
TCP does
2. It is possible that the NetEnforcer will be able to differentiate
between P2P UDP and VoIP UDP, but probably not in the real time - 1-2
seconds - that will be needed
3. If an ISP attempts to throttle UDP, it is very likely that VoIP
traffic will be adversely affected.
4. UDP is 'all or nothing', that is, it doesn't have the ability to be
gracefully controlled like TCP. If a UDP packet is dropped there is no
handshake exchange at the protocol level. So it may reduce the CPU load
on both the NE and out routers.
5. It also remains to be seen just what effect trying to rate limit UDP
traffic will have on VoIP. At the moment only about 5% of our traffic
is UDP. But that will change to 50%+. VoIP can stand much more packet
loss than data, but I don't know what effect rate limiting say 1Gbps of
UDP to 800Mbps or 500Mbps might have on the very small part of it that
it will have. It may be very little, if, say a queuing algorithm like
Weighted Fair Queuing is used.
It may even be the case that a simple rate limit on UDP traffic, in
conjuction with WFQ will allow us to effectively control P2P traffic if
it is all UDP. In that case, we would no longer need the NetEnforcer.
I think there is some, but not a great, chance that will be the case.
6. BUT, such throttling will only affect traffic once our routers or
the NetEnforcer have processed it. It will not reduce the traffic
hitting out network. Unless our bandwidth suppliers, and all the worlds
networks, are able to limit UDP in a similar way we will no longer be
able to control the traffic coming to us.
With TCP, it is possible to reduce the traffic rate from just one point
in the network because information can be sent back to the origin to
tell the server to reduce the rate it is sending. UDP has no such
controls and the server will just keep sending information as fast as it
can or as fast as the application tells it to.
7. One benefit is that if all P2P traffic is UDP, it will make it
easier for the PeerApp to identify and should improve our cache
efficiency a bit.
8. I think it will affect other networks much more than ours - very
major networks too. So the backlash may be sufficient from Tier 1
carriers that Bittorrent decide not to use UDP after all."
Doubtless Steve will refine his views when he isn't responding to such queries at around midnight.
So, it remains to be seen whether or not BT et alia will switch to using UDP and if that does happen what the actual impact on the global internet is. If it is anything like what the article's author describes and what Steve's first pass view describes then one of four things will happen:
1) The internet will collapse under the 'weight' of a bunch of unethical people's activities.
2) P2P (UDP or not) will be banned on every network that has the capability of doing that - which means every commercial network.
3) Some new method of switching data will be 'invented' that will allow legitimate traffic to be processed without delay.
4) The cost of internet services will increase and be purely based on pay per megabyte downloaded and uploaded (no "included download" plans)
I think that 4) above, which is pretty much the way Telstra, and some others, currently charges retail customers will become the standard method of providing internet services so every current user can thank the copyright pirates for their future higher internet pricing - should that be the case.
By luck or good judgment (take your pick) Exetel have spent over two years taking the pain and grief of managing a long and large volume 'free period', understanding P2P controlling, understanding P2P caching and, more recently, basing plans on PAYU as well as, in the case of HSPA, charging for uploads as well as downloads. Whatever experience we have gained from these very painful learning curves may well be more useful in the future than it appeared when we began to understand how we might need to manage a large network at some future point in time.
Sometimes planning pays off - even if it turns out not to be quite what you were planning for.