Can you recall a time when you went to a friend’s new place for dinner without consulting your smartphone? How about a time when you drove across state lines without looking at an app for directions? You probably cannot recall when that happened. Navigation systems today are fast, accurate, and reliable. They provide rich information on traffic, provide convenient voice prompts, tell us about locations, including ratings of those places, and even guide us when we make mistakes in following instructions.
The reason navigation systems have become indispensable is because they provide us with the right information when required. They don’t tell us about the traffic in London when we are trying to get from home to an office in Boston. They give us the information we need to make accurate navigation decisions.
Internet traffic has very different characteristics today compared to a decade ago. Traffic patterns have changed and traffic volumes have grown drastically. Our dependence on networks for finding faster routes has increased with real-time applications like autonomous cars and remote-assisted surgeries. This makes it imperative to have an efficient navigation system for networks. Wouldn’t it be great if each router could have a navigation system of its own? Much like the navigation systems in our smartphones and cars, these routers would be able to figure out the best paths depending on where packets need to go and localized real-time information on the pathways.
Unfortunately, legacy routing protocols that establish connectivity for most of the Internet today do not convey real-time traffic or load information. This makes it impossible to make sound decisions for navigating through network pathways efficiently. Many companies like Facebook, Google, and others have built their own systems to convey real-time information on capacity and performance of pathways to make these decisions. Some have even eliminated traditional routing protocols within their networks. However, there are significant challenges to replacing these routing protocols on a massive scale globally. Any solution must work with existing protocols.
Traditional networking vendors have continued to invent additional protocols or extend legacy protocols to add service information or network state information. These have not found wide acceptance as they increase complexity and make networks difficult to troubleshoot. These methods also do not take into account modern developments in other fields like machine learning, artificial intelligence, big data, and high-speed databases. Handheld or car navigating systems are successful and effective because they go outside the boundaries of traditional maps and provide real-time actionable data.
There is a dire need for routers to be able to exchange service information, which is analogous to having information on locations. It’s also important for routers to exchange connection information to those services, which is analogous to traffic information to get to those locations. This information should only be shared when there are travelers or packets interested in going to those locations. Networks must perform the following functions to support modern cloud, IoT, and 5G applications:
- Service Discovery: A method for routers to dynamically learn where services are located similar to how users search for Walmart or Target on their phones.
- Multi-hop path selection: Routers should be able to calculate paths to these services based on real-time network loads and path performances. This is analogous to finding the best route to the local Walmart depending on road and traffic conditions.
- Analytics and heuristics: By analyzing network performance information over time, network elements can predict future delays or loads similar to how Google maps can tell you how much time it will take to get to Walmart tomorrow at a given time.
For the foreseeable future, existing routing protocols will continue providing connectivity information across the internet. While businesses have become much more connected and service-centric over the past decades, there has been little innovation to the network. The network has only become more complex with the addition of tunnels, firewalls, load balancers and deep packet inspection – each a “band-aid” approach that doesn’t address the underlying issues. What’s really needed is innovation to the router itself.
Systems need to enable routers to learn services and path performance information, allowing them to choose pathways efficiently, just like modern navigation systems do. Next time you feel the need to whip out your smartphone and look up directions to a location, keep in mind that the routers in modern networks can benefit from a similar app too.