To cope with an ever increasing number of internet connected devices, large scale computer clusters are becoming an everyday requirement for any web-service provider; and with the prevalence of compute clouds, these can be obtained effortlessly at a scale that was previously unimaginable. However, the distribution models available have not caught up with the advancements of clouds yet, and as such, distributed programs running on top of these platforms require significant efforts to take full advantage of their hosts' capabilities.
We introduce a unique concept into the domain of distributed computing, namely, that the idea of a specific node is impedimental to scalability. The proposed model to achieve a robust system is to raise the unit of abstraction from a single node to a cluster of nodes sharing the same responsibility; and to eliminate all notions related to location, further breaking dependencies between distributed entities.
Iris is a decentralized messaging framework with the specific goal of simplifying the development and maintenance of cloud based back-end services by striving towards zero configuration through complete self-organization; achieving full messaging security out of the box; whilst providing the communication primitives to support easily assemblable distributed systems.
Speakers: Péter Szilágyi