|Abstract||As demonstrated by the emergence of paradigms like fog computing  or cloud-of-clouds , the landscape of third-party computation is moving beyond straightforward single datacenter-based cloud computing. However, building applications that execute efficiently across data-centers and clouds is tedious due to the variety of communication abstractions provided, and variations in latencies within and between datacenters.
The publish/subscribe paradigm seems like an adequate abstraction for supporting “cross-cloud” communication as it abstracts low-level communication and addressing and supports many-to-many communication between publishers and subscribers, of which one-to-one or one-to-many addressing can be viewed as special cases. In particular, content-based publish/subscribe (CPS) provides an expressive abstraction that matches well with the key-value pair model of many established cloud storage and computing systems, and decentralized overlay-based CPS implementations scale up well. On the flip side, such CPS systems perform poorly at small scale. This holds especially for multi-send scenarios which we refer to as entourages that range from a channel between a publisher and a single subscriber to a broadcast between a publisher and a handful of subscribers. These scenarios are common in datacenter computing, where cheap hardware is exploited for parallelism (efficiency) and redundancy (fault-tolerance).
In this paper, we present Atmosphere, a CPS system for cross-cloud communication that can dynamically identify entourages of publishers and corresponding subscribers, taking geographical constraints into account. Atmosphere connects publishers with their entourages through überlays, enabling low latency communication. We describe three case studies of systems that employ Atmosphere as communication framework, illustrating that Atmosphere can be utilized to considerably improve cross-cloud communication efficiency.|