Microsoft Technology Licensing, LLC

Peter A Christofferson

Brian Thomas Berkowitz

Gunnar Mein

Sreenivas Narayana Simhadri

Each node in a network accessing a common database caches a copy of the database tables it uses and updates those tables based on changes made by other nodes in a coherent cache, i.e., those node that also have a copy of the tables in their cache. When a new node joins the coherent cache, dynamic synchronization of the table is performed by having one of the existing nodes act as a source to populate the table(s) required on the new (destination) node from its own copy. The dynamic synchronization controls the population operation and allows for the populating of tables on the destination node from partially populated tables on the source node. The dynamic synchronization of the tables permits database updates to continue to be made to the tables by the existing nodes in the coherent cache and provides a mechanism to propagate the committed updates to the tables on the destination node. Once the tables on the destination node are fully populated, all pending updates are suspended while a snapshot of the state of the table on the source node is taken. Once the destination node has processed the snapshot, active transactions on the source node are initiated on the destination node, and all transactions are allowed to modify the table on the destination node.

Dynamic synchronization of tables

An arbitration process ensures changes made by more than one node to the same data are propagated to each node holding the shared entity and applied in the same order on each node. An arbitration cycle for a particular entity is begun on a node ('instigator') when the node broadcasts a proposal message or when the node ('observer') receives such a proposal message. Multiple nodes can be instigators during a single arbitration cycle. Each node that receives a proposal message sends a response message to the corresponding instigator. After each instigator node receives all the proposals in the arbitration cycle, it determines whether it is the winner of the arbitration cycle and broadcasts a closure message if it is. Each node determines an order in which to apply the changes when it has received all the proposals. Arbitration cycles associated with different items can be running concurrently.

Arbitration of state changes

When multiple nodes in a computer network perform work on a shared entity, such as a database record, an arbitration process is used to ensure changes made by more than one node to the same data is propagated to each of the nodes holding the shared entity and that the changes are applied in the same order on each of the nodes. An arbitration cycle for a particular entity is begun on a node ('instigator') when the node broadcasts a proposal message for a change to the entity to the other nodes or when the node ('observer') receives such a proposal message. Multiple nodes can be instigators during a single arbitration cycle. Each node that receives a proposal message sends a response message to the corresponding instigator. The arbitration cycle is closed to additional proposal messages after one instigator has received responses from all the nodes that share the entity. After each instigator node receives all the proposals in the arbitration cycle, it determines whether it is the winner of the arbitration cycle and broadcasts a closure message if it is. Each node determines an order in which to apply the changes proposed in the arbitration cycle when it knows it has received all the proposals in the arbitration cycle as determined by the sending or receiving of the closure message. Because each arbitration cycle is associated with a particular entity, multiple arbitration cycles associated with different items can be running concurrently in the network but multiple arbitration cycles for the same item must be executed serially. The arbitration process is executed even if there is only one proposal made in an arbitration cycle since the arbitration process enables the orderly propagation of changes across all the nodes that share the entity.

When multiple nodes in a computer network perform work on a shared entity, an arbitration process is used to ensure changes made by more than one node to the same data is propagated to each of the nodes holding the shared entity and that the changes are applied in the same order on each of the nodes. An arbitration cycle for a particular entity is begun on a node ('instigator') when the node broadcasts a proposal message for a change to the entity to the other nodes or when the node ('observer') receives such a proposal message. Multiple nodes can be instigators during a single arbitration cycle. Each node that receives a proposal message sends a response message to the corresponding instigator. The arbitration cycle is closed to additional proposal messages after one instigator has received responses from all the nodes that share the entity. After each instigator node receives all the proposals in the arbitration cycle, it determines whether it is the winner of the arbitration cycle and broadcasts a closure message if it is. Each node determines an order in which to apply the changes proposed in the arbitration cycle when it knows it has received all the proposals in the arbitration cycle as determined by the sending or receiving of the closure message.

Methods and systems for managing state changes during an arbitration cycle when multiple computer nodes request changes of shared data

Clearing share latches for databases, such as in-memory databases, is disclosed. In one embodiment, a computerized system includes a database having a number of sections, a client process and a core process. The client process is capable of asserting a share latch on a section of the database. The core process is capable of asserting an exclusive latch on a section of the database. The core process is further capable of clearing the share latch asserted by the client process, upon termination of the client process while the share latch is still asserted. Specific methods for such clearing, as well as latch logs, are also disclosed.

Share latch clearing

When multiple nodes in a computer network perform work on a shared entity, such as a database record, an arbitration process is used to ensure changes made by more than one node to the same data is propagated to each of the nodes holding the shared entity and that the changes are applied in the same order on each of the nodes. An arbitration cycle for a particular entity is begun on a node (&ldquo;instigator&rdquo;) when the node broadcasts a proposal message for a change to the entity to the other nodes or when the node (&ldquo;observer&rdquo;) receives such a proposal message. Multiple nodes can be instigators during a single arbitration cycle. Each node that receives a proposal message sends a response message to the corresponding instigator. The arbitration cycle is closed to additional proposal messages after one instigator has received responses from all the nodes that share the entity. After each instigator node receives all the proposals in the arbitration cycle, it determines whether it is the winner of the arbitration cycle and broadcasts a closure message if it is. Each node determines an order in which to apply the changes proposed in the arbitration cycle when it knows it has received all the proposals in the arbitration cycle as determined by the sending or receiving of the closure message. Because each arbitration cycle is associated with a particular entity, multiple arbitration cycles associated with different items can be running concurrently in the network but multiple arbitration cycles for the same item must be executed serially. The arbitration process is executed even if there is only one proposal made in an arbitration cycle since the arbitration process enables the orderly propagation of changes across all the nodes that share the entity.

An in-memory database system uses a shared memory to cache records and keys read from a database and controls the updating of the records and keys through a database manager process. When a transaction performs an update, the original, unmodified data is preserved in the shared memory, the new data is written to the shared memory, and a look-aside table for the transaction records the changes. A transaction performs read-only access to the shared memory using its own context while a versioning scheme based on the look-aside tables ensures a read-committed isolation level view of the original, unmodified data until the modifying transaction commits the update. The database manager is responsible for writing the new data into the shared memory and for maintaining the look-aside tables for all transaction which have made modifications to the data in the shared memory. The database manager also writes committed changes to the database and performs rollback on uncommitted changes in the shared memory using the entries in the look-aside table for the committing/aborting transaction. The shared memory is divided into logical pages and short duration page latches are employed to maintain consistency on the page while a transaction or the database manager is reading or writing data on the page.

Growing community of inventors

Kenmore, WA, United States of America

Peter A Christofferson