Scaling Instant Messaging Communication Services: A Comparison of Blocking and Non-Blocking Techniques

1. Introduction

Instant Messaging (IM) and presence services have become a mainstay of modern communications. Consumer messaging services such as Windows Live Messenger (n.d.), Google Talk (n.d.), and AOL Messenger (http://nodejs.org/).

This last approach (node.js) has achieved some surprising results recently, particularly in addressing the well known C10k problem (Kegel, 2011). Put succinctly, this C10k problem names a limitation of most web servers: they can handle at most 10,000 connections simultaneously. Node.js approaches are showing some interesting results when applied to this problem (Salihefendic, 2010). This work is not quite a replication of the C10k problem: we are more interested in a messaging and presence services than a plain HTTP service (which is the focus of most C10k experiments). However, we believe that we have conducted some interesting experiments in devising a hybrid environment where high volume processing is now possible, even in the context of interacting with a more traditional “blocking” service such as Openfire.