Many spiders make a better web: a unified web-based actor framework
Florian Myter, Christophe Scholliers, Wolfgang De Meuter · 2016
JavaScript is the predominant language when it comes to developing applications for the web. Many of today's web-based systems are implemented solely in JavaScript due to its applicability to both client and server-side development. This use of JavaScript as a general purpose programming language has sparked a number of efforts to provide programmers with the means to write parallel web applications. For the most parts these efforts have based themselves on actor-based parallelism, given the similarities between actors and the JavaScript runtime model. We argue that these solutions are based on actor models which do not optimally fit web development. Concretely, existing solutions fail to provide programmers with an actor framework which embraces both parallelism and distribution. To this end we present Spiders.js, an actor framework providing both high-level parallelism and built-in distribution. In Spiders.js, programmers can easily specify the coarse-grained parallelism needs of modern web applications. Moreover, Spiders.js' built-in distribution features allow programmers to express client/server, server/server and client/client distribution simply by using actors. We show the performance characteristics of our approach by detailing Spiders.js' results for the Savina benchmark suite.