Sap Se

Matthias Hauck

Hannes Jakschitsch

Gerald Franz

Roland Sedler

Umang Rawat

Christian Bensberg

Daniel Ritter

Romans Kasperovics

Markus Fath

Stefano Belloni

Alan Braenzel

Frank Tetzel

Mathias Kemeter

Marc Jacob

The present invention relates to an improved database system architecture designed for generating and managing heterogeneous graphs. The system utilizes a document store in conjunction with a persistent adjacency index to enhance flexibility, adaptability, and transactional visibility during the generation and modification of complex graph structures. Data stored in the document store utilizes a flexible schema for storing data which improves the efficiency in both storing and retrieving data from the document store during graph-related processing tasks.

Graph processing using a document store

A method may include generating, for a heterogeneous graph comprising a plurality of vertex tables and/or edge tables, an adjacency structure including a vertex adjacency structure specifying a plurality of neighboring vertices in the heterogeneous graph and an edge adjacency structure specifying a plurality of neighboring edges in the heterogeneous graph. A graph processing algorithm operating on the heterogeneous graph may be executed based on the adjacency structure. Related systems and computer program products are also provided.

Adjacency structure for heterogeneous graph data

A method may include receiving a workspace definition defining a graph workspace for a heterogeneous graph having a plurality of vertex tables and/or edge tables. Upon validating the workspace definition, the graph workspace for the heterogeneous graph may be created in accordance with the workspace definition. The graph workspace may include a label for each table of the plurality of vertex tables and/or edge tables comprising the heterogeneous graph. Each vertex and edge of the heterogeneous graph may be uniquely identified based on a label of a corresponding table and one or more key attributes. A graph processing algorithm operating on the heterogeneous graph may be executed based on the graph workspace. Related systems and computer program products are also provided.

Graph workspace for heterogeneous graph data

The present disclosure involves systems, software, and computer implemented methods for configuring graph query parallelism for high system throughput. One example method includes receiving a query to be executed against a graph database. System properties are determined of a system in which the query is to be executed. Algorithmic properties are determined of at least one algorithm to be used to execute the query. Graph data statistics are determined for the graph database. Graph traversal estimations are determined for a first iteration of the graph query and an estimated cost model is determined for the first iteration based on the graph traversal estimations. Estimated thread boundaries are determined for performing parallel execution of the first iteration. Work packages of vertices to be processed during the execution of the first iteration are determined based on the first estimated cost model and the work packages are provided to a work package scheduler.

Configuring graph query parallelism for high system throughput

A system, a method, and computer program product for performing buffering operations. A data update is received at a buffering location. The buffering location includes a first buffer portion and a second buffer portion. The data update includes an address tag. The buffering location is communicatively coupled to a memory location configured to receive the data update. A target address of the data update in the memory location is determined using the first buffer portion and compared to the address tag. The data update is applied using the first buffer portion to update data in the first buffer portion upon determination that the target address matches the address tag. The target address of the data update is pre-fetched from the memory location upon determination that the target address does not match the address tag. The first and second buffer portions buffer the data update using the pre-fetched target address.

Buffering of associative operations on random memory addresses

Growing community of inventors

Frankfurt-am-Main, Germany

Matthias Hauck