Other

Amadeus S.a.s.

Gustavo Alonso

Guy Pardon

Donald Kossmann

Dietmar Fauser

Georgios Giannikis

Cedric Florimond

Philipp Unterbrunner

Jeremy Meyer

Cédric Florimond

A distributed data store includes a storage layer comprised of a plurality of storage nodes, each storage node configured to receive and queue table operations, and an aggregation layer comprised of aggregator nodes configured to route table operations to storage nodes in the storage layer and merge results of table operations returned by storage nodes in the storage layer. Each storage node includes in part scan threads dedicated to continuous and full scanning of segments of a table that are respectively associated with processor cores upon which such scan threads are running.

Continuous full scan data store table and distributed data store featuring predictable answer time for unpredictable workload

A method for storing and retrieving data in a storage node of a data store and storage node of a data store, storing in main-memory at least one segment of a relational table. The storage node includes at least one computational core running at least one scan thread each dedicated to the scanning of one of the at least one segment. At least one scan thread uniquely, continuously and exhaustively scans the dedicated segment of the relational table. The storage node receives and processes batches of query and update operations for the at least one segment of the relational table. The query and update operations of a batch are re-indexed at beginning of each scan by the scan thread. Then, the indexed query and update operations of a batch are independently joined to data records of said segment that match with predicates of the indexed query and update operations.

In a light-weight transactional system architecture, each component providing a service performs its own scheduling and transaction management. An invocation of one of these services (through remote method invocation) results in the creation of a local transaction (child of the invoking transaction and parent of any transaction that might be triggered by invoking the services of other servers). The information that is required to build a global composite transaction is implicitly added by the system to each call. Each transaction is, however, independently handled at each server. That is, the servers neither communicate among themselves nor rely on a centralized component to make scheduling or recovery decisions. In this way, components can be dynamically added or removed from the system without compromising correctness. Regardless of the configuration, the system according to the invention guarantees that transactions executed over these servers are correct and recoverable at a global and local level.

Decentralized, distributed internet data management

A light-weight architecture is provided, where each component is in itself its own advanced mini-transaction processing monitor. To accomplish this, the system is most readily implemented as a set of Java classes. The resulting architecture is as follows. In a composite system, each server is an independent component performing its own scheduling and transaction management. These servers are built using Java and inheriting from the classes provided by the system according to the invention. The interface to each server defines the services it implements. An invocation of one of these services (through remote method invocation) results in the creation of a local transaction (child of the invoking transaction and parent of any transaction that might be triggered by invoking the services of other servers). Each transaction is a thread that (in an exemplary system) can invoke SQL statements in a local database (directly connected to that server) as well as services offered by other servers. All the information that is required to build a global composite transaction is implicitly added by the system to each call. Each transaction is, however, independently handled at each server. That is, the servers neither communicate among themselves nor rely on a centralized component to make scheduling or recovery decisions. In this way, components can be dynamically added or removed from the system without compromising correctness. All a new server needs to know is the interface and address of the servers it will invoke. Regardless of the configuration, the system according to the invention guarantees that transactions executed over these servers will be correct (serializable) and recoverable at a global and local level.

Growing community of inventors

Zurich, Switzerland

Gustavo Alonso