A Realization Method of Load Balancing Network Middleware Based on Extensible Protocol

Abstract

The invention discloses a load balancing network middleware implementation method based on an extensible protocol: 1. The server receives the multi-frame message from the client; Step 2. After parsing the multi-frame message into a single-frame message, the message is obtained type; step 3, put the read message into the read thread pool, and put the write message into the write thread pool, wherein, the small size message is allocated on the stack, and the large size message is allocated on the heap; step 4, the read thread pool and the The write thread pool calls the load balancing algorithm to select specific processing threads, and puts the message content into the corresponding thread message queue; step 5, for the write message, the write thread writes the data to the disk file and returns a confirmation message; for the read message, and the reading thread queries the corresponding disk file according to the index, reads the data into the cache, and assembles the message and returns it to the corresponding client. With the advantages of high performance, scalable protocol and load balancing, it can effectively save network traffic and improve response rate.

Description

technical field [0001] The invention relates to a method for implementing a load balancing network middleware based on an extensible protocol. Background technique [0002] Time-series data processing technology is widely used in wide area measurement system (WAMS), substation monitoring, dispatching, DC, stability control and other power automation systems to meet the needs of storing massive and high-frequency data. Time-series databases are mainly used to process high-frequency data at the second and millisecond levels. Each storage cycle (milliseconds or seconds) will generate a batch of data, so the amount of data is huge, and the read and write performance of network traffic and network middleware is extremely demanding. High, this requirement is different from traditional network middleware. It is necessary to customize and develop network middleware suitable for time-series databases, and optimize it for message, network and storage characteristics. [0003] Traditi...

AI Technical Summary

Problems solved by technology

First of all, these frameworks are general-purpose frameworks, and the key-value pair characteristics for time-series data cannot be optimized, which makes the communication protocol too large and wastes resources; second, the general framework needs to consider too many scenarios, so the middleware is too complicated, such as CORBA. A heavyweight framework will drag down the deployment of the entire time-series database; again, traditional middleware shields the underlying operating system dependencies very well, achieving versatility and cross-platform, but this is based on the loss of certain system performance, so Unable to meet the requirements of large data volume and low response delay of time series data

Images