Neural-network-based hot spot data prediction method and system

Abstract

The invention provides a neural-network-based hot spot data prediction method. The neural-network-based hot spot data prediction method includes the steps: acquiring the historical access record in the preset time quantum for a user as a supervising data set; taking the supervising data set as the training data, and training a multi-layer neural network model until the accuracy of predicting the hot spot data by the multi-layer neural network model is greater than a threshold to obtain a target multi-layer neural network model; and finally traversing multiple data stored in a storage system, and based on the target multi-layer neural network model, successively determining whether each data stored in the storage system is hot spot data in the next preset time unit. According to the historical access record of the user, the neural-network-based hot spot data prediction method effectively predicts the data that the user may access in the future, loads the hot spot data of the next time unit in advance, and unloads the non hot spot data of the next time unit in the cache in advance, thus improving the using experience of the user and improving the system performance.

Description

technical field [0001] The present invention relates to the technical field of data storage, and more specifically, to a method and system for predicting hotspot data based on a neural network. Background technique [0002] In modern storage systems, both cache loading and hotspot data statistics have a lag. Generally, data is loaded into the cache only after it has been accessed once or multiple times, and then it is known that this is a data hotspot, and then it is accelerated to load, affecting The overall performance and user experience of the system are improved. [0003] In addition, some old hot data has not released the cache for a long time. These old hot data were loaded into the cache because they were frequently accessed in the past time period, but in the current and future time periods, these old hot data are no longer frequently accessed but still Stored in the cache, resulting in cache congestion, new hot data cannot be added to the cache, affecting the impr...

AI Technical Summary

Problems solved by technology

[0002] In modern storage systems, both cache loading and hotspot data statistics have a lag. Generally, data is loaded into the cache only after it has been accessed once or multiple times, and then it is known that this is a data hotspot, and then it is accelerated to load, affecting Overall system performance and user experience

[0003] In addition, some old hot data has not released the cache for a long time. These old hot data were loaded into the cache because they were frequently accessed in the past time period, but in the current and future time periods, these old hot data are no longer frequently accessed but still Stored in the cache, resulting in cache congestion, new hot data cannot be added to the cache, affecting the improvement of system performance

Images