Argentina (University of Palermo) A key for the internet to continue growing

Alejandro Popovsky, the Dean of the Faculty of Engineering at the University of Palermo, said that UP developed an algorithm that allows overcoming the biggest obstacles to traffic through the network.

A key for the internet to continue growing

The University of Palermo developed an algorithm that allows overcoming the biggest obstacles to traffic through the network, which grows with the evolution of technology and the greater number of users.

Congestion control is responsible for the fact that the internet has been successful as a technology and that it grew from a few nodes to a global network of billions of nodes. We could say that congestion control used on the internet is one of humanity’s most successful technologies. Traditional congestion control aims at two goals: (A) maximize bit-rate, (B) minimize data packet loss.

But it totally neglects the following aspects: (C) minimize the delay between the sender and receiver of the data and (D) that there is a fair distribution of the available capacity of the network. The objective of the fair distribution of available network capacity is easily understood. But the goal of minimizing delay is not so clear: if I maximize the transmission speed, why am I interested in minimizing the delay? The answer is that in many cases communication is transactional, that is, it consists only of very short query messages, and their corresponding very short response messages, and the total time of these transactions is almost independent of the bit-rate between the ends, but depends almost entirely on the delay between them.

So, if congestion control generates large delays for all those connections that share a network bottleneck with the first connection, all transactional communications get much worse. This includes web browsing, which performs a transaction for every small graphic or script included on the page. For the user who is browsing or using services through the Internet, the quality of his experience depends only on the total waiting time for each of his actions or requirements, and does not care if this time is caused by a decrease in transmission speed or if it is caused by delays somewhere on the Internet where the communication passes. This user usually shares part of the internet path with other users in the same household or in the same company. And he even shares with himself with the many transmissions that are generated from his computer or cell phone, about which many times he does not even have notion. So a single communication that induces large delays in the bottleneck will generate a decrease in the quality of the experience of the same user and the other users with whom he shares.

However, the four objectives mentioned above do not depend solely on the congestion algorithm, which runs on computers at both ends of the connection, but can also be affected by the dynamics of the device that hosts the communication bottleneck. To improve this dynamic, intelligent packet outbound queue (AQM) management algorithms can be incorporated into these devices, to induce data packet senders to efficiently use available capacity without flooding bottleneck buffers, and to fairly distribute their capacity.

So, depending on who is addressing the problem of communications quality and performance in the service, is where you can act:* If it is internet service providers or hosting or cloud service providers, they can act through the choice of the congestion control algorithm on the servers, or through the application of AQM in the routing devices and also the firewall of the datacenter (very rarely contemplated).* While users can act through the selection of the congestion control algorithm on your computers or through the incorporation of AQM in your connection devices (the minifirewall or modem that connects them to the Internet). * In the case of the end user, who is mostly a consumer of content, you need a congestion control algorithm for incoming traffic, of which there are very few options being the Palermo congestion control, technology developed by the University of Palermo, an option. But to work the problem from your network connection device, the provider does not give you permissions to do so.

There is, however, the option of adding in series to the connection device a second device that limits the transmission speed in a value a little lower than the contracted, but that allows it to become the bottleneck. This allows you to intelligently apply the bottleneck packet queue. The devices (router or home firewall) that allow this configuration are usually called devices with SQM (smart queue management).

SQM has the disadvantage that the capacity available for incoming traffic is not constant, as internet providers over-book (oversell their capacity). So the available capacity usually varies greatly with the hour, and even minute by minute. Variation that routers and home firewalls with SQM cannot follow. While a good congestion control algorithm for incoming traffic constantly adapts to these changes. But on the other hand, it is not simple for end users to install this option on their operating system. In the case of Internet access service providers, the complication usually comes because the bottleneck is usually in the device that applies the limitation to the bandwidth contracted to all users, where it is very difficult in general, to combine its functionality with AQM.

The UP has created the Palermo congestion control algorithm, which combines the four objectives: bit-rate maximization, packet loss minimization, delay minimization and fair capacity distribution. And it can be applied both to improve outbound traffic on servers (sender side congestion control), and to improve inbound traffic (receiver side congestion control) on users’ computers or on web proxies. The outbound traffic version is available in open source mode for all Linux distributions. There are other very good options also such as BBR congestion control, but it exists only for outbound traffic control.

The technologies that made possible the revolution of the internet and communications continue to evolve as services grow in performance and data volume, while the number of users and smart devices grows, and users grow in their dependence and demand. of quality. We will be able to build on the success of congestion control technologies to keep this growth possible. *Dean of the Faculty of Engineering of the University of Palermo.