Cisco just released their latest CRS-3 internet backbone router. The box supports an aggregate throughput of 322 Tbit/s. Wow ! A capacity never heard of. This indicates some underlying trends. Let’s look at this.
On one hand, this indicates that the capacity of the connections increases steadily. When I started -a long time ago- we used to offer “Ethernet” LAN cards at speeds of 1 Mbit/s … These speeds were later increased to a strange 2.5 Mbit/s, then 5 and finally the regular 10 Mbit/s. Now we are talking about speeds of 40 Gbit/s and 1000 Gbit/s with fiber optic connections. The new CRS-3 has 100 Gbit/s fiber connections. Multiply this by the number of ports and you’ll see why we need throughput capacities of 322 Tbit/s …
On the other hand, Cisco states that the internet has moved from a message based system to a real-time video system. YouTube has 1 billion video downloads a day. We will get HD videoconferencing and TV over the internet. The internet is undergoing a change towards a world we did not anticipate in the beginning : moving from text to full multimedia.
Thirdly, with the advent of higher connection capacities, we’ll see even more internet utilisation. Today there is the discussion about moving internet connections for the home to 100 Mbit/s. Suppose there are 4 million homes in Belgium, with each a 100 Mbit/s connection. 4 M x 100 Mbit/s x2 (full duplex) = 800 Tbit/s. This means we would need 3 of these Cisco CRS-3 to deliver a full symmetrical 100 Mbit/s service from any home to any home. Wow ! 3 of these boxes would be sufficient to build the internet for the whole of Belgium !
You might think “yeah … will we use this capacity ?” Probably not in the whole time and probably not fully symmetrical. Watching TV requires more downstream capacity than upstream, for example. But then you have the traffic peaks, the multiple devices at home, the new (multimedia) applications. We will have moments where we need the full 100 Mbit/s (at peak), and the operators will have to build high capacity networks to absorb these peaks.
The next step is Petabits per second throughputs, but that’s probably for in 5 years 
Most Comments