Gettys has been writing about bufferbloat recently, something that "wrecks latency" and saps network performance no matter how fat your network pipe happens to be. This isn't exactly a new concept. Gettys points to Stuart Cheshire wrote about the problems with latency in 1996 in "It's the latency, Stupid" which debunks the idea that more network capacity will always be faster:
Bandwidth is a measure of capacity, not a measure of how fast the network can respond. You pick up the phone to send a message to Shanghai immediately, but dispatching a cargo ship full of blu-ray disks will be amazingly slower than the telephone call, even though the bandwidth of the ship is billions and billions of times larger than the telephone line. So more bandwidth is better only if its latency (speed) meets your needs. More of what you don’t need is useless. Bufferbloat destroys the speed we really need.
Gettys has been writing about the problem of "bufferbloat" on his blog for a few months, and says that the problem is similar to "frogs in slowly heating water." According to Gettys, individual problems with bufferbloat have been detected, but "we've been engineering around this problem without full understanding, by throwing bandwidth at the problem and as gamer routers [routers targeted at gamers to reduce lag in network games] show. But RAM costs have dropped even faster than network speeds have risen, and rising network speeds encourage yet larger (currently unmanaged) buffers; throwing bandwidth at the problem has been a losing race."
Worse, Windows XP is being retired. Wait, what? That's a good thing, right? Not necessarily. Gettys says that XP doesn't implement TCP window scaling, which means that XP isn't contributing (as much) to bufferbloat as modern OSes that do. As XP slowly retires Gettys says "dominant TCP traffic will shift from partially able to saturate most Internet links to always capable of link saturation."
So now that the problem is (at least partially) understood, Gettys has launched Bufferbloat.net to try to tackle the problem. The initiative includes several mailing lists for discussion and development, and his presentation given recently at Bell Labs about the problem.
It will be interesting to see whether Gettys work is going to gather significant interest. So far it seems like a fairly small group of folks on the mailing list, but one hopes they will find some usable solutions and drive more interest in tackling the problem.
