“Those that can bend with the wind will weather the storm” (unattributed)
It is no surprise that social distancing induced by Covid-19 is driving a rise in Internet use. Some of the recent headlines:
Yet, the majority of these articles go on to say the Internet will be “okay”, that it is built to sustain spikes in traffic. While spikes typically are expected to be of short duration, say around a large sporting event, this is just a long drawn out spike. Business as usual.
I think it is safe to say that personal experience bears this out. On top of work at home, we have all been using the Internet more for personal interactions, such as for Zoom lectures or exercise classes. While sometimes we may complain about the chopiness of a connection, for the greater part, everything works.
Can you imagine what the situation would have been a mere 30 years ago? The Internet was still in its infancy and most of our communication was voice telephony. This relied on, can you imagine, dedicated circuits. With everyone trying to call everyone, we would have needed to wait hours to get a dial tone!
The genius of the packet-switched Internet is that it is able to make near complete use of communications channels by simultaneously multiplexing multiple streams of traffic together statistically. In other words, the channel is “shared” so everyone has access.
A popular mechanism for achieving this is Weighted Fair Queuing, illustrated here.
The situation is that there are multiple streams of services traffic (in the diagram three streams) all aggregating at a network node and contending to be forwarded to a downstream node in the network.
In the diagram, a frame selector (packets are organized into frames) processes ‘up to’ 50 frames from service 3. This value of 50 is the buffer weight. When 50 frames are processed, or when there are no more frames to process, the selector processes ‘up to’ 30 frames for service 2, followed by ‘up to’ 20 frames from service 1, and then the cycle repeats.
By adjusting the size of the buffer weights, it is possible to change the relative performance for each service type. By dynamically adjusting the weightings, we can try to meet the relative performance expectations for all services as best as possible.
The WFQ approach underlies the “best effort” service commitment that accompanies most Internet services. It has multiple benefits. It is simple to implement, it can be adjusted dynamically to deal with surges in traffic from different sources, and it ensures maximum use of the egress pipe to forward Internet traffic. And that, more or less, is how the Internet bends when network traffic grows from normal wind conditions into a storm.
Jonathan Homa, is Senior Director Portfolio Marketing at ECI for the last few years. Jonathan enjoys French beer in Nice and loves to discuss the 3 hottest topics in networking today: softwarization, virtualization and cloudification.