MIT gets hit again! The problem with slow pages is known from the beginning of the Internet. Too many companies, researchers and developers have dealt with its solution but no one has ever managed to announce a 34% success rate, reducing the total page load time. Until today.
A team of researchers from the Department of Computer Science and Artificial Intelligence Lab (CSAIL) of MIT achieved impressive results.
The MIT team, including also a Harvard professor, has created the Polaris algorithm, which focuses on determining the right time to launch network requests to Web resources or Web resources.
The websites work in a very simple way. For someone to have access to one by clicking here types a URL to program his tour. A DNS server redirects it to the IP address where the website is hosted and the browser starts downloading the files (Web resources) stored on the server, say an HTML page.
Within this HTML page, there is the source code of the web page that loads different resources in CSS format, JS files, images, Flash content, and other information. Each resource of these means a separate request, which (applications) makes the user's browser.
As previous studies have shown, the problem of slow page loading is not always due to the small bandwidth, but also to the time it takes to send all network requests, the size of downloaded files, and the delays of the network itself .
How does the Polaris algorithm work?
To address these issues, MIT's Polaris framework will work by creating dependency graphs for each page on the Web, which will dictate a more efficient order in which all the resources of a page should be loaded.
Charts are widely used in development today software, and are at the heart of some software development tools.
In fact, the Polaris framework works first as it records how the loads of a web page interact with each other. It then creates a dependency graph for each WEP page, and sends the requests in such a way that only the necessary content is loaded first.
MIT has tested Polaris in 200 different network conditions and found that load times decreased by an average of 34%. The best results were achieved on larger websites that contained many JS files.
The researchers will present the Polaris framework at the USENIX Symposium to be held this week at Networked Systems Design and Implementation.