A method and an application of multi-process single-write multi-read lock-free shared memory

Abstract

Aiming at the problems that multiple processes simultaneously read and write the same shared memory data area, the data is frequently updated, and the kernel lock causes serious performance loss, theinvention provides a multi-process single-write multi-read lockless shared memory method and application, which avoids the use of locks through data structure design and atomic operation. When the data is updated, a piece of unused shared memory space is re-applied from the shared memory for the data to be updated, and the data is written into the memory space and the data key value relationship is modified; At the same time, the presence of the old data area ensures that it is recycled when no thread reads, Memory allocation for the next new or updated data, As that key-value relationship isupdate only after the data is update, any data area is not read and written at the same time, and the use of the lock can be completely avoided. The invention fully utilizes the atomic operation characteristic and completely avoids the use of the lock, thereby realizing the real-time data access capability with high performance when the data is updated at high frequency. The shared memory data access performance is improved.

Description

technical field [0001] The present invention relates to a fast trading method for brokers based on a memory database, specifically a lock-free shared memory method in a multi-process single-write-multiple-read scenario, which is applied to real-time high-frequency data between multiple processes through shared memory technology Shared scenarios can be used as a technology for designing high-performance key-value in-memory databases. Background technique [0002] As my country's securities market matures, the proportion of professional investors such as institutions in the securities market will also increase. In mature stock markets such as Hong Kong and the United States, more than 70% of the transaction volume is completed by institutional investors. With the continuous growth of institutional clients and high-end clients of securities companies, the increasing number of programmatic transactions, the demand for fast and safe high-end channels is also becoming more and mor...

AI Technical Summary

Problems solved by technology

The key performance factor for data sharing between multiple processes through shared memory is the data synchronization mechanism. However, most in-memory databases currently improve data access performance by designing efficient locking mechanisms. These solutions can indeed meet the needs of most application scenarios. , but for real-time data updated frequently, its performance will still be greatly affected due to the existence of lock conflicts

In scenarios where data is frequently updated, frequently copying the data in the current data area to the backup area will cost a lot, especially when the amount of data is large

Images