Abstract: A database system uses byte ordering for keys and a trie index to reference stored data. The keys of a database are converted into byte-comparable sequences of byte values. The trie index is generated including nodes connected by edges defining paths from a root node to leaf nodes. Each edge is associated with at least one byte value such that each path from the root node to a leaf node through one or more edges defines a unique byte prefix for a byte-comparable sequence of byte values. The leaf node of each path is associated with a database location value. A record is accessed in the database using a database location value determined from referencing the trie index using a byte-comparable sequence of byte values of the record generated from a key of the record. A trie structure and byte ordered keys may be used for partition or row indices.

Abstract: In partitioning a graph database, a plurality of vertices of the graph database is assigned to a plurality of nodes. The vertices of the graph database are connected by edges that indicate relationships between the vertices. A vertex of the graph database is designated as a super-vertex that is split into a truncated vertex and at least one vertex representative.

Abstract: A database system uses byte ordering for keys and a trie index to reference stored data. The keys of a database are converted into byte-comparable sequences of byte values. The trie index is generated including nodes connected by edges defining paths from a root node to leaf nodes. Each edge is associated with at least one byte value such that each path from the root node to a leaf node through one or more edges defines a unique byte prefix for a byte-comparable sequence of byte values. The leaf node of each path is associated with a database location value. A record is accessed in the database using a database location value determined from referencing the trie index using a byte-comparable sequence of byte values of the record generated from a key of the record. A trie structure and byte ordered keys may be used for partition or row indices.

Abstract: Embodiments relate to a compacting datafiles generated by a database node using a compaction processing node with separate compute resources. The database node generates datafiles and stores the datafiles in a data store. To perform compacting of the datafiles, a snapshot of the data store is created and stored in a snapshot store separate from the data store. The compaction processing node is initiated and attached with the snapshot store. The compaction processing node generates a compacted datafile that is stored in the snapshot store. The database node replaces the data store with the snapshot store, and writes additional datafiles using the snapshot store as a new data store. The compaction processing node may be an instance of a cloud compute infrastructure that is initiated to perform the compaction to reduce compute resource usage by the database node.

Abstract: A distributed system that manages resources of the distributed system without the need for complex time synchronization systems is described. The distributed system includes a resource manager that manages the resources of the distributed system. The resource manager assigns leases and renews leases of resources of the distributed system to clients in the distributed system. The leases specify the duration of time that the lease is awarded to clients.

Abstract: Data consistency across replicas in a cluster of nodes is maintained by continuously validating local data ranges and repairing any inconsistencies found. Local data ranges are split into segments and prioritized. After a segment is selected for validation, a hash value of a portion of the segment is compared to a hash value from other nodes storing replicas of that data. If the hash values match then the data is consistent. If the hash values do not match then the data is not consistent and whichever data is most current according to their timestamps is considered correct. If the local node data is correct, it is communicated to the replica nodes so they can be updated. If the local node data is not correct, then data from the replica nodes is correct and is used to update the data in the local node. An alternative, incremental validation approach improves efficiency.

Abstract: Systems and methods are described herein for efficiently updating a secondary index associated with a log-structured merge-tree (LSM) database. A Global approximate member query (AMQ) Filter is queried to determine whether a primary key, retrieved from a list of LSM database updates, already exists in the LSM database. If the primary key does not already exist in the LSM database then read-before-write and delete operations, typically performed with known approaches, do not need to be performed on the secondary index in order to update the secondary index, thereby avoiding significant additional computer processing and input/output operations.

Abstract: At least a portion of a graph database having a plurality of vertex-centric indices is stored. A virtual edge to be generated is identified based on a plurality of edges of the graph database. The virtual edge connecting at least a pair of vertices that were not previously directly connected is generated. The plurality of vertex-centric indices is updated to include information about the virtual edge.

Abstract: In partitioning a graph database, a plurality of vertices of the graph database is assigned to a plurality of nodes. The vertices of the graph database are connected by edges that indicate relationships between the vertices. One or more abstract paths between one or more vertices of the graph database are identified. Each abstract path is weighted based on a likelihood of a database query following the abstract path. The vertices of the graph database are assigned to the nodes according to the abstract paths between the vertices.

Abstract: Data consistency across replicas in a cluster of nodes is maintained by continuously validating local data ranges and repairing any inconsistencies found. Local data ranges are split into segments and prioritized. After a segment is selected for validation, a hash value of a portion of the segment is compared to a hash value from other nodes storing replicas of that data. If the hash values match then the data is consistent. If the hash values do not match then the data is not consistent and whichever data is most current according to their timestamps is considered correct. If the local node data is correct, it is communicated to the replica nodes so they can be updated. If the local node data is not correct, then data from the replica nodes is correct and is used to update the data in the local node.

Abstract: In partitioning a graph database, a plurality of vertices of the graph database is assigned to a plurality of nodes. The vertices of the graph database are connected by edges that indicate relationships between the vertices. A vertex of the graph database is designated as a super-vertex that is split into a truncated vertex and at least one vertex representative.

Abstract: Various operations, functionalities and systems are described herein for backing up one or more node to an offsite location, restoring the one or more node from the offsite location, restoring the one or more node to a point-in-time (PIT) from the offsite location, cloning the one or more node from the offsite location, and cloning the one or more node to a PIT from the offsite location. Example operating contexts include one or more cluster of nodes running a NoSQL (Not only Structured Query Language) distributed database and backup, restore and/or cloning on those one or more cluster of nodes.

Abstract: Various operations, functionalities and systems are described herein for backing up one or more node to an offsite location, restoring the one or more node from the offsite location, restoring the one or more node to a point-in-time (PIT) from the offsite location, cloning the one or more node from the offsite location, and cloning the one or more node to a PIT from the offsite location. Example operating contexts include one or more cluster of nodes running a NoSQL (Not only Structured Query Language) distributed database and backup, restore and/or cloning on those one or more cluster of nodes.

Abstract: Fault tolerant querying of data distributed across multiple nodes is accomplished by each node determining and reporting its own health status and indexing status to the other nodes in the cluster via a gossip protocol. A coordinator node then prioritizes replica nodes based on the received status of the other nodes and sends query requests to those nodes based on the prioritization. Should a node fail to provide an response to a query request, further query requests are sent to a next highest priority replica node containing the relevant data. This results in improved query performance by avoiding busy nodes and further provides a fault tolerant approach to data queries.

Abstract: A distributed system that manages resources of the distributed system without the need for complex time synchronization systems is described. The distributed system includes a resource manager that manages the resources of the distributed system. The resource manager assigns leases and renews leases of resources of the distributed system to clients in the distributed system. The leases specify the duration of time that the lease is awarded to clients.

Abstract: Described embodiments provide systems and methods for building client server applications. The application server code is deployed within a distributed datastore and utilizes the datastore's data locality information to fulfill requests with minimal remote procedure calls (RPC), reduced transfer of data between servers, and shared data caches. The application server inherits the properties of the distributed datastore such as statefulness, replication and failover. Described embodiments also enable the fetching and processing of data in a “big data” environment—a system that contains multiple servers' worth of data, with improved time and bandwidth considerations.

Abstract: Described embodiments provide systems and methods for building client server applications. The application server code is deployed within a distributed datastore and utilizes the datastore's data locality information to fulfill requests with minimal remote procedure calls (RPC), reduced transfer of data between servers, and shared data caches. The application server inherits the properties of the distributed datastore such as statefulness, replication and failover. Described embodiments also enable the fetching and processing of data in a “big data” environment—a system that contains multiple servers' worth of data, with improved time and bandwidth considerations.