Microsoft Technology Licensing, LLC

SungJu Cho

Andrew Joseph Carter

Jane Alam Jan

Qingpeng Niu

Sanjay Sachdev

Roman A Averbukh

Yinyi Wang

Scott M Meyer

Andrew Rodriguez

Karan Rajen Parikh

Joshua D Ehrlich

Shyam Shankar

Jiahong Zhu

Yanwei Zhang

The disclosed embodiments provide a system for managing a materialized view of data in a data store. During operation, the system obtains a first creation time of a first materialized view of the data in the data store. Next, the system verifies that the first materialized view is not affected by deletions applied to the data since the first creation time. The system then identifies one or more additions to the data that are not included in the first materialized view. Finally, the system applies the additions to the first materialized view to maintain an eventual consistency of the first materialized view with the data store.

Managing materialized views in eventually consistent distributed data stores

The disclosed embodiments provide a system for processing data. During operation, the system obtains a group of two or more integer values. Next, the system sets, for each integer value in the two or more integer values, a delta encoding tag that represents a delta encoding of the integer value with respect to at least one other integer value in the group and uses a delta encoding state represented by the delta encoding tag to calculate an encoded value from the integer value. The system then encodes a length of the encoded value in a length tag for the integer value. Finally, the system replaces the integer value in the group with an encoded version containing the delta encoding tag, the length tag, and the encoded value.

Encoding edges in graph databases

The disclosed embodiments provide a system for updating an index structure of a graph database storing a graph. During operation, the system includes, in the index structure, a first compressed edge store containing a first compact representation of edges in the graph at a first virtual time and a first series of updates to the edges after the first virtual time. At a second virtual time, the system creates a second compact representation of the edges from the first compact representation and the first series of updates. The system then appends, to the second compact representation, a second series of updates to the edges after the second virtual time to produce a second compressed edge store. Finally, the system updates the index structure by atomically replacing, in the index structure, a reference to the first compressed edge store with a reference to the second compressed edge store.

Atomic updating of graph database index structures

The disclosed embodiments provide a system for processing data. During operation, the system obtains a graph of a social network, wherein the graph includes a set of nodes representing users in the social network and a set of edges representing relationships between pairs of the users. Next, the system stores, on a single computer system, a static representation of the graph, wherein the static representation includes a first set of fixed-size blocks representing the nodes and the edges and a first index that maps a set of identifiers for the nodes and the edges to offsets of the first set of fixed-size blocks. The system then uses the static representation of the graph to process, by the single computer system, one or more queries of the graph.

Fast querying of social network data

The disclosed embodiments provide a system for processing data. During operation, the system determines a current incoming queries per second (QPS) to one or more components for processing queries of a graph database, wherein the graph database is replicated across multiple clusters and distributed among a set of storage nodes in each of the clusters. Next, the system uses the current incoming QPS to estimate, for the one or more components, an expected QPS associated with fanning out of the queries to the clusters. The system then selects a number of clusters in the multiple clusters for fanning out of a query based on the expected QPS and one or more throughput limits for the one or more components. Finally, the system transmits the query to one or more of the storage nodes in the selected number of clusters.

Throughput-based fan-out control in scalable distributed data stores

A social graph is divided into a series of partial graphs having limited hops and reference counts. For each of a plurality of nodes in the social graph, a partial graph for the node is created having a first degree list of nodes reachable in one hop and a second degree list of nodes reachable in two hops. Reference counts of how many paths exist between the node and each node reachable in two hops are also added to the second degree list. A global inverted index is maintained containing a list of all nodes in the social graph and for each node in the social graph a list of nodes that directly connect to the node. The partial graphs created for each of the plurality of nodes are distributed across a plurality of data servers. An index of which partial graphs are stored on which data servers is maintained.

Growing community of inventors

Cupertino, CA, United States of America

SungJu Cho