Key-key-value stores for efficiently processing graph data in the cloud

Modern cloud data storage services have powerful capabilities for data-sets that can be indexed by a single key-key-value stores-and for data-sets that are characterized by multiple attributes (such as Google´s BigTable). These data stores have non-ideal overheads, however, when graph data needs to be maintained; overheads are incurred because related (by graph edges) keys are managed in physically different host machines. We propose a new distributed data-storage paradigm, the key-key-value store, which will extend the key-value model and significantly reduce these overheads by storing related keys in the same place. We provide a high-level description of our proposed system for storing large-scale, highly interconnected graph data - such as social networks - as well as an analysis of our key-key-value system in relation to existing work. In this paper, we show how our novel data organization paradigm will facilitate improved levels of QoS in large graph data stores.