CNS Core: Small: Fast or Dynamic Websites? Eliminating the Need to Choose

To support evolving Internet services such as social networking, healthcare, and education, web pages increasingly embed large amounts of dynamic and personalized content. However, doing so renders previously routine recommendations for ensuring low-latency page loads insufficient, e.g., content delivery networks and caches that focus on static content.

As a result, developers today face a challenging trade-off between serving rich, dynamic content and ensuring the fast page loads that their users expect, both of which greatly affect user experience. The goal of the proposed work is to holistically rethink how page loads are executed in order to address this tension, without sacrificing page correctness or end-to-end security.

The key insight underlying the proposed work is to eschew optimizations for dynamic content that are siloed at either the client- or server-end of page loads, and instead pursue optimizations that synergize operations at both locations. The project involves three applications of this approach that collectively address the delays associated with dynamic content throughout the page load process.

First, it develops novel page template strategies to mask the blocking delays for dynamic content fetches at the start of page loads (e.g., HTML). Second, it introduces a suite of new techniques for accelerating the execution of JavaScript code -- the primary source of intra-load dynamism -- via efficient and state-preserving parallelism and computation reuse.

Third, it develops new page serving architectures that efficiently generate dynamic content close to clients and maximize utility of the client’s browser cache.

The proposed research aims to democratize the ability to serve state-of-the-art web pages with high performance, thereby targeting a wide slice of the developer ecosystem. In particular, small content providers today lack the expensive infrastructure that their larger counterparts leverage to serve dynamic content with low latency; the result is a sacrifice of either performance or functionality, both of which present an uphill battle in user acquisition and retention.

The proposed systems can enable performant (but affordable) dynamic content using existing web infrastructure. The project also involves outreach efforts to attract students from populations currently under-represented in Computer Science. Key to these efforts is the relatability of the web, which eliminates the terminology barrier and lets students develop services similar to those that they use daily, increasing interest.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.