Algorithms for building highly scalable distributed data storages
The ability to scale is a desirable business requirement for computer systems. Data storage is a central part of any computer system. The scalability of any computer system strongly depends on the ability to scale of its data storage. One of possible ways to implement highly scalable data storage is to build a structured P2P network on the level of data or on the level of physical nodes. This approach has been successfully used in well-known DHT systems, which can be considered as distributed key-value storages. However the search in DHT systems is restricted by the search of exact hash values. This makes it impossible to build data storages that can efficiently perform complex queries. In this talk we will discus a class of algorithms, which overcomes this limitations and can be used as a basis for building highly scalable distributed storages with the ability to efficiently perform similarity search in metric and semimetric spaces.