33results about How to "Improve concurrency efficiency" patented technology

The invention discloses a distributed transaction processing method and system. The method includes: starting the outermost transaction, and assigning a unique outermost transaction ID; calling the RPC interface of the multi-service configuration service to allocate resources for different services, and The outer transaction ID is passed to the multi-service configuration service; the outer transaction ID is used as the parent transaction to nest multiple layers of new transaction IDs in turn, and the new transaction ID of each layer is the child transaction of the previous layer; from the parent transaction to the child transaction direction, in turn Perform resource allocation for multi-service configuration services, network element services, and basic services on each new transaction ID, and upload the corresponding service name and parent transaction ID; when all resource allocation operations in all transactions are successful, the user clicks on the GUI to submit the configuration , call Commit of the transaction manager to save the entire transaction. The invention ensures the final consistency of data, avoids the use of global locks, and greatly improves the concurrency efficiency of user operations.

The application aims to provide a cache management method and device. Compared with the prior art, the cache management device provided by the application queries and determines a target cache record matched with main key information in a cache containing a plurality of cache records, wherein each cache record comprises corresponding citation information and latest access information; and updating the citation information and the latest access information corresponding to the target cache record. The citation information and the latest access information are set in the cache records, so that the cache management device can be helped to set a replacement mechanism of the cache records; the cache management device ensures that the cache records active recently are stored in the cache, so that the active cache records are prevented from being replaced; and meanwhile, "the citation information" and "the latest access information" can be modified through only one atomic instruction without the need for performing a lock operation on the cache records, so that global lock contention during modification of an existing cache management algorithm is avoided and the concurrent efficiency is effectively improved.

The invention relates to the technical field of data storage, and provides a method, device and server for orderly controlling and storing information by lock-free threads. The method includes: according to the non-release state of the shared storage node in the circular queue, calling the CAS instruction during the period when the focus thread in the at least two first threads stores the current data in the shared storage node; The release state is changed to the release state, and the release state is used to indicate that the second thread can read the current data; the second thread is allocated a shared storage node in the release state; according to the storage position of the shared storage node in the circular queue and the original cursor in the CAS instruction Consistency, update the original cursor with a single vector, so that the focus thread locks the shared storage node. In this way, the threads can be controlled to access data in an orderly manner, which overcomes the additional overhead of controlling the storage queue and improves the thread concurrency efficiency.