An automatic approach for creating faster websites using HTTP/2, machine learning, and AI.
Horizon 2020 SME Instrument, proposal ID 868576
Website speed is the key to success for e-commerce websites. A fast website creates a better user experience, improves Google ranking, and most importantly, it increases conversion. Slow websites cost millions of dollars a year in lost sales, and therefore this is a pressing problem on a growing market that call for efficient solutions.
As web usage and online commerce has exploded, we have moved from simple static websites to fully interactive and complex websites, introducing new challenges for creating fast websites. For example, loading a website is more re- source intensive, websites are loaded using smartphones in wireless networks, and website visitors are spread globally. To address these new challenges of the web, a new communication protocol, HTTP/2, was released in 2015. The new protocol does not require any changes to existing websites, and by default it provides a more efficient loading of resources, creates more mobile-friendly sites, and improves Google rank. The new protocol also introduces new features that software can take advantage of to substantially increase website speed.
However, most of the HTTP/2 features are not straightforward to apply and need to be actively configured, and no standards or guidelines exist to inform developers on optimal usage. This leave the problem for the developers to do manual tuning, which is both cumbersome and might demand changes every time the website is updated. Since the features can lead to substantial speed improvements, the challenge of adopting them has introduced a major business opportunity:
The main objective of this project is to develop the software product ShimmerCat Accelerator from TRL stage 6 to TRL stage 7. This main objective can be divided into two subobjectives.
Goal: Prove the technical viability of the product by demonstrating it in an online operational e-commerce environment that takes live web traffic
Goal: Prove the commercial viability of the product by involving users that in the future can work as resellers.
Goal: Compile the feasibility report.
To prove the technical viability of ShimmerCat, we have configured it to serve an online e-commerce environment that takes live web traffic. For comparison, the same website has been setup with both ShimmerCat and Nginx.
E-commerce served with traditional Nginx web server
E-commerce served with ShimmerCat web server
To prove the commercial viability of the product we have been able to involve users that in the future can work as resellers.
The project results will be presented in a Feasibility report that will be sent to the H2020 Agency.
Website speed is the key to success for e-commerce websites. A fast website creates a better user experience, improves Google ranking, and increases conversion. Optimizing slow websites is therefore a pressing problem on a growing market that calls for efficient solutions.
As web usage and online commerce has exploded, we have moved from simple static websites to fully interactive and complex websites, introducing new challenges for creating fast websites. To address these new challenges, the web protocol HTTP/2 was released in 2015. The new protocol introduces new features that software can use to substantially increase website speed.
However, most of the HTTP/2 features are not straightforward to apply and need to be actively configured. Since the features can lead to substantial speed improvements, the overall objective of this project was to develop a software product that can utilize the features of HTTP/2 to help online businesses create faster websites.
The project outcome shows that the automatic approach with ShimmerCat for creating faster loading times, that uses machine learning and HTTP/2, has the potential to significantly improve the performance of e-commerce websites.
The work performed during the project has focused on exploring and assessing the technical feasibility and commercial viability of the software product ShimmerCat, a web server that handles requests for things which are part of a website. During the project ShimmerCat has been developed to send we traffic data to a cloud infrastructure that analyzes the data, and sends back directives about how to manage the traffic to improve website speed.
During the project the technical viability of the product has been validated in an online operational e-commerce environment that takes live web traffic. The project has also resulted in a validation of the products commercial viability, thanks to the help of an ecommerce platform that has implemented the product in order to resell it.
ShimmerCat goes beyond the state of the art as it can collect, analyze, and optimize performance with data from HTTP headers and other metadata (timestamp, TCP details, TLS and protocol fingerprints, etc). This data is analyzed to automatically build rules for HTTP/2 Push and resource preloading, rules for redirecting automatic traffic, above-the-fold image prioritization data, and next-gen optimized image formats. The potential impact of the project is to be able to create websites that load 30-60% faster.
If you want to know more about the project, don't hesitate to get in touch!