Wednesday, April 8, 2015

Gigabit Speeds Don't Improve Experience, Content Delivery Networks and Caching Does

According to a study by Mike Belshe, “if users double their bandwidth without reducing their Round Trip Time (RTT), the effect on Web browsing will be a minimal improvement (approximately five percent).”


“However, decreasing RTT, regardless of current bandwidth always helps make web browsing faster,” Belshe argues.


Faster local Internet access connections do help, up to a point. After about 10 Mbps, no single user is likely to see much improvement, if at all, in page load times, for example. The U.S. Federal Communications Commission and U.K. Ofcom agree: beyond 10 Mbps per user, experience is not measurably improved--if at all--by faster Internet access speeds.


Bandwidth (in Mb/s)
Page Load Time via HTTP
1
3106
2
1950
3
1632
4
1496
5
1443
6
1406
7
1388
8
1379
9
1368
10
1360


Although there is a considerable jump in the early bandwidth speed increases, the returns as the pipe gets bigger continue to diminish until they are almost negligible.


The important observation is that the measure of a digital experience isn’t just--or primarily--about the speed of download.


Latency, or round trip delay, is more fundamental, beyond a minimum amount of access speed.

That is one reason so many large application providers use content delivery networks that place content closer to end user locations, in principle improving round trip delay.

Shockingly, and all marketing claims notwithstanding, end user experience of Internet apps is primarily a matter of latency, not access bandwidth. That is because "non-network" sources of delay generally represent an order of magnitude more impairment than local access speed, or even all network delay, taken together.

No comments:

"Tokens" are the New "FLOPS," "MIPS" or "Gbps"

Modern computing has some virtually-universal reference metrics. For Gemini 1.5 and other large language models, tokens are a basic measure...