Huawei Technologies Co., Limited

Huaxin Zhang

Yuk Kuen Chan

Wenbin Ma

Xun Xue

Ronen Grosman

Emad Boctor

Ping Chen

Sherman Lau

Hoi Leng Lao

Rongzhen Cui

Method, apparatus and system for assigning a commit sequence number (CSN) to a WRITE transaction in a network having nodes and a global time server. The CSN is defined by a timestamp of the WRITE transaction and an error bound of the timestamp. The WRITE transaction is committed after the timestamp is issued and an amount of time equal to the error bound plus a time adjust value has passed. The time adjust value is based on round-trip times between the plurality of nodes and the global time server. The time adjust value may be the longest expected round-trip time. By waiting for an amount of time equal to the error bound plus the time adjust value, any READ transaction occurring after the WRITE transaction may receive a READ timestamp without any delay, provided the READ timestamp error bound is less than or equal to the time adjust value.

Global clock overhead with asymmetric waiting time

A system and method for an online transaction system to transition from operating under a first mode (FM) to operating under a second mode (SM), and vice versa. In the FM, transaction nodes identify transactions with sequential numbers obtained from a global transaction server (GTS). In the SM, transaction nodes identify consecutive transactions with timestamps based on global clock servers (GCSs). When transitioning from one mode to the other mode, a controller instructs the transactions nodes to provide GCS-based transaction timestamps and error bounds to the GTS, which modifies the timestamps and the error bounds based on previously received timestamps and error bounds, and on the latest transaction number issued by the GTS. Modified timestamps and error bounds are obtained at the transaction nodes. Each transaction is timestamped with its respective modified timestamp. But there is a delay equal to the modified error bound before requesting another modified timestamp.

Systems and methods for a distributed database to transition between centralized transaction management and global clock transaction management with little to no downtime

Methods and apparatus are described for splitting a page (or node) in an index tree database index for a multi-master database. The methods include creating a first split page of a first page. The first page belongs to a first set of pages and comprises a hinge page. The hinge page also belongs to a second set of pages. Updating the first set of pages may cause the hinge page itself to split, creating a split of the hinge page. The second set of pages is updated in response to creating the split of the hinge page. Described are techniques whereby the updates to the first and second sets of pages are atomic, that is, they all succeed or they all fail, which may realize any of several benefits.

Method and apparatus for consistent page split operation in a multi-master database

The disclosed systems and methods are directed for detecting and resolving write-write conflicts among a plurality of transactions received from master nodes of a multi-writer database system. The method includes receiving a plurality of REDO logs and storing the plurality of REDO logs in a buffer, each REDO log associated with the one of the plurality of transactions, selecting one REDO log of the plurality of REDO logs; persisting the transaction associated with the one REDO log in a local storage when a write-write conflict is detected between the one REDO log and at least one other REDO log of the plurality of REDO logs prior to committing the transaction associated with the one REDO log; and transmitting a status of the transaction associated with the one REDO log to a global transaction manager (GTM).

Method and system for detecting and resolving a write conflict

Generating, by a first primary site that is included in a group of primary sites of a distributed database system, a commit action redo log message for a commit action performed by the first primary site for a first database transaction, the commit action redo log message including a transaction ID for the first database transaction and a transaction order indicator that represents an order of database transactions in a group of database transactions performed at one or more primary sites of the group of primary sites; and sending, by the first primary site, the commit action redo log message, for a corresponding first standby site that backs up the first primary site.

Distributed database remote backup

The disclosed systems, structures, and methods are directed towards for processing a lock request by a master node to access a shared data object in a database located at a remote location comprising identifying, in the master node, a category of the lock request. If the lock request is for a self-compatible lock, determining a value of a counter associated with the master node and with one or more other master nodes, and if the value of the counter is zero, locally granting the lock request. If the value of the counter is greater than zero, forwarding the lock request to a global lock manager (GLM), and if the lock request is for a heavy-conflicting lock, forwarding the lock request by the master node to the GLM.

Method and system for fast processing of locks requested to access a shared resource

The disclosed computing node comprises a processor and a non-transitory storage medium storing instructions executable by the processor. A method and a system are also disclosed. A subset of a plurality of conventional redo records, corresponding to received write requests, is selected based on an identical data location identifier. The conventional redo records of such selected subset are combined into a consolidated redo record. The consolidated redo record is then transmitted to a target node for processing.

Growing community of inventors

Markham, Canada

Huaxin Zhang