August 7th, 2005
The next ten years in network is going to be 10x more exciting than the last ten we have. We had our first billion internet users finally after nearly three decade but over the next ten years, we going to add another (or two) billion more easily.
What’s significant is the additional billion users will not be geeks but just ordinary user. Yep, geeks like us will be outnumbered! And to them, Internet is just a tool, like a phone and the uses of Internet is far more important then the technology driving it. We already seeing more attention have been given to the applications (bittorrent, skype, blogs) and less into the core networking technology (ipv6, ipsec) in the last couple of years.
Yet the core networking technology is more important than ever despite the lack of attention especially the core Internet is facing sever problems.First, the incumbents are starting to understand all-IP network, or what they called NGN and that to me is a bad news. Already operators are talking about ‘using IP but not public Internet’ or providing an ‘complete triple-play experience’. With the additional billion of users, who don’t really care if they have freedom to connect, there is a very good opportunity in the “users are dumb” business just like the good old days of POTS. Good for telcos, bad for the Internet, who is build on the concept of “End-to-End”, not dumb users.
Like it or not, they control the network and the billions end-users is going to outnumber us. Our demand for a clean IP pipe may not be heard. Already, I am facing difficulties in getting my ISP to stop filtering my ICMP. Sooner or later, I might even get a pipe with SIP port filtered off all in the name of ‘providing consistent quality service’.
From a technical perspective, we are already pushing IP to the limits.
BGP routing problem is a serious issues that still does not have any solution within the IETF – we just keeping our finger crossed that not too many small network asking for multihoming, or we tell those who wants it that it cannot be done (yea sure. haha).
Then we are putting IP over every infrastructure we can think of – wifi, wimax, pons. Yea, why not? But already, we seeing wireless (layer 2) people finding the need to integrate with IP more closely because of the high error rate on wireless. And anyone remember Vint Cerf’s experiment to do interplanetary Internet (yep, no kidding! they put IP on probes we send out in space)? Imaging how well TCP/IP is going to function with a round-trip latency of, oh say, 24 hours? (Most of us would be screaming if the latency is more then 1000ms).
TCP is also facing several limitations. Ever try transfering a big file, say 1gb over a 622mbps pipe over a long distance? Simple layman calculation say it takes almost two minutes but in realty, you get only a few mbps out of the pipe using TCP (depending on the distance/latency) due to what we called “big fat pipe” problem. So even we have 20mbps broadband or even fiber to the home in the next few years, don’t dream you can get HDTV from across the globe, just not yet.
All these are problems and I am sure it is solvable in one way or another. e.g. P2P like BitTorrent helps to bypass (but not overcome) the “big fat pipe” problem. But it is like putting band-aids on IP and then hoping it wont fail us.
So far, the 30 year-old technology has work quite well for us. But in the next ten years, where we expect 2-3 billions users with an always-on 1gb broadband, 100mbps mobility connectivity, a trillion devices on huge interplantary network across earth, moon and perhaps mars, I began to doubt IP is the only answer.