Computer scientists working at the MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) recently developed an algorithm called Polaris that could boost browsers’ load speed by 34 percent.
Researchers noted that their approach was completely different from previous similar attempts. Other teams have tried to improve page load times by simply improving data compression of elements on a website.
Polaris, on the other hand, tries to create a map of the relationships (also known as ‘dependencies) between different objects on a website to boost speed. If the browser has a map of those relationships beforehand it can choose the shortest route to load the content of a complex webpage.
It is not the first time a team tries to track dependencies to improve speed. But unlike previous attempts, MIT team’s latest approach is more “fine grained” as it can detect subtle dependencies that other technologies have failed to track.
Ravi Netravali, a MIT researcher involved in the project, explained that previous attempts to map dependencies were linear. Today when a user clicks on a web page the browser loads a specific list of items on that page in a specific order. So, if one item is listed ahead of another, the browser assumes that this is the order to load the items.
Netravali noted that this approach does not reflect the real dependencies between items, and it is incredibly slow. Polaris, on the other hand, can track which object interacts with another object on a page i.e. an object could write something and the other object could read. That is a true dependency. If two objects have nothing to do with one another, there isn’t a dependency, and Polaris treats them separately.
Researchers also explained that existing browsers make a lot of round-trips between objects on a website which considerably slows down load times. Though it takes only 100 milliseconds to load a specific object, if the webpage is bloated with content the load times can add up.
Plus though current browsers are heavily optimized, the round-trip times can slow down page load speed, which is extremely inconvenient especially on mobile devices. Netravali noted that desktop browsing is much cheaper than mobile browsing since the cost of accessing a network is much lower on desktop than on a cell network. Past research had found that every extra millisecond in load times can make content providers lose both revenue and users.
Image Source: Pixabay