Local Transaction Data Copies for Distributed Grid Throughput
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed data grids face reduced throughput due to the need for locks on transaction data, which limits efficiency and increases the number of remote calls when using pessimistic or hybrid locking approaches, even with optimistic locking, as data grids maintain copies of transaction data in the same location as the actual data.
Innovation Solution
Implementing a method where a transaction originator process in a data grid maintains a local copy of remote transaction data, allowing it to perform operations without acquiring locks on the remote data until the transaction is prepared, thereby minimizing the use of distribution algorithms and network calls.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pessimistic or hybrid locking approach is used in data grid, then data consistency is maintained, but database throughput is greatly reduced
Solution Approach 1:
The patent creates local copies of remote transaction data in the transaction originator process's data structure. This copying mechanism allows the system to operate on local data without acquiring locks on remote data during the transaction scope, eliminating the throughput bottleneck caused by pessimistic locking while maintaining consistency through version comparison at commit time.
Solution Approach 2:
The patent performs preliminary actions by copying remote transaction data to local data structures before the transaction begins. This preliminary copying eliminates the need for continuous remote data access and lock acquisitions during transaction execution, thereby improving throughput while maintaining data consistency through subsequent version verification.
2Productivity
If optimistic locking is implemented in data grid, then throughput is improved, but efficiency decreases due to increased remote calls
Solution Approach 1:
The patent stores copies of remote transaction data locally in the transaction originator process's data structure. This local copying eliminates the need for repeated remote calls during transaction execution, reducing network overhead and improving efficiency while maintaining the benefits of optimistic locking for throughput.
3Reliability
If locks are acquired during transaction operations, then data consistency is ensured, but transaction speed is reduced due to lock acquisition waits
Solution Approach 1:
The patent creates local copies of remote transaction data that can be accessed and modified without acquiring locks. This copying mechanism allows transactions to proceed at full speed by operating on local data, while data consistency is maintained through version number comparison and optimistic locking verification at commit time.
Solution Approach 2:
The patent performs preliminary data copying before transaction execution, which eliminates the need for lock acquisition during transaction operations. This preliminary action allows transactions to execute at maximum speed using local data, with consistency checks performed only at commit time.
4Measurement precision
If remote transaction data is accessed repeatedly in transaction scope, then accurate data is used, but network overhead increases due to frequent remote calls
Solution Approach 1:
The patent stores local copies of remote transaction data in the transaction originator process's data structure. This copying eliminates repeated remote calls during transaction execution, significantly reducing network overhead. Data accuracy is maintained through version number tracking that verifies the local copy remains consistent with the remote data.
Data Source
AI summary
A first process executing in a computer system in a data grid receives a request to perform at least one transaction operation of a transaction based on remote transaction data of at least one other process in the data grid. The at least one other process is capable to perform a prepare operation for a transaction. The first process identifies a local copy of the remote transaction data in a local data structure and performs the at least one transaction operation using the local copy of the remote transaction data without acquiring a lock on the remote transaction data.


