Today, the most common traffic on the internet is multimedia traffic. This traffic accounts for 70% of the total internet traffic, with the main streaming sites taking the lion’s share of that slice of the pie That is why a few years ago, the IETF asked itself, “How can we optimize the management of multimedia traffic? And the answer to that question was ALTO.
ALTO (Application-Layer Traffic Optimization) is a network protocol that seeks to expose the current state of the transport network to multimedia resource distribution systems (MDS), which are responsible for storing and distributing resources such as a live broadcast of a football match, the new season of your favorite series or the live stream of a streamer from twitch. Knowing how to distribute this information so that it arrives as quickly as possible to all the users of the network and at the same time do it with the least possible number of replicas is a problem of resource allocation that cannot be solved only with the knowledge of how saturated the caches and servers of the MDS are, it is also necessary to know where the requests come from and what capacities we have available to send the resources to all the users who request them.
This is where ALTO comes in, exposing the capabilities of the transport network so that applications can make decisions on how to optimize the management of their network resources so that when you turn on Netflix to watch the new season of your favorite series, it arrives as quickly as possible to your home.
How can ALTO help the NEMO project?
This problem that MDSs face with content distribution also occurs in other contexts, such as distributed networked systems. In these systems we have a very clear optimization challenge: when and where should we create a new instance of the system? To solve this question, at NEMO we have thought of an architecture that evaluates the current and future state of the network through technologies such as Machine Learning or Digital Twins, but to obtain this vision of the transport part we still need something to obtain these metrics and provide that missing piece for our puzzle; and that is where ALTO has fitted in.