Research On Distributed Lock Technology

With the rapid development of computer technology, has now entered the era of cloudcomputing. In the era of cloud computing, most of the services are deployed in a distributedenvironment. In a distributed environment, many services are to allow some failures, but alsoallow data inconsistency. But some of the most basic services is needed to remain highreliability and consistency, these services are based distributed system to provide normalservice.In multiple backup systems in order to improve data access efficiency and reliability ofthe data, there is often more than one data redundancy, and is distributed on differentmachines. In order to maintain the consistency of the data, some concurrency control methodsneed to be applied. In distributed environment, concurrency control methods include lock,timestamp and multi-version concurrency control method. In the scenes which the writeoperation is large probability of conflict and re-execute the transaction cost of larger, the lockmethod approach is more appropriate. This article focused on this scenario, to studydistributed lock implementations, made a number of design:1. We propose a distributed lock system architecture, which is highly available and faultrecovery time in second-level. In the paper, mainly through3points to complete.First, designed and implement a lock data storage and lock logic separation systemarchitecture; Secondly, client implemented consistency Hash to achieve loadbalancing of lock requests; Thirdly, using Redis master-salve solution to store thelock data, this solution also improves the performance of Redis and solve Redis’ssingle point of failure problem; In the mean while, using Redis hot backupautomatically switch solution to shorten recovery time after a fault.Overall, thesolution also avoids single point of failure of the system and improve the scalabilityof overall system.2. We Provide a efficient client implementations.3. The client implemented deadlock prevention function.4. In a locked using the first-come first-served mechanism to achieve a fair schedulingstrategy to avoid starvation problem.5. In the service side, implementing a memory pool to improve memory allocationefficiency.