Method and device for storing time series data in internal storage

Abstract

The invention provides a method and device for storing time series data in internal storage. The method comprises the steps that time series query information is obtained; based on the time series query information, corresponding time-bias information is determined, wherein the time-bias information comprises hour information, minute information and bias information; based on the time-bias information, leaf nodes on a corresponding time tree are searched for; the leaf nodes are searched for corresponding time series data based on the bias information. The method and device have the advantages that the method has low complexity and improves the query efficiency of the time series data; furthermore, the device occupies less internal storage and supports out-of-order writing.

Description

technical field [0001] The invention relates to the technical field of computers, in particular to a method and device for accessing time series data in memory. Background technique [0002] With the development of computer and Internet technology, time series databases are more and more widely used. A time series database is a database that stores massive time series data (that is, each data corresponds to information at a time point), and can support the query of these massive data. [0003] One of the main usage scenarios of time-series databases is monitoring and query. Users will frequently query the data of the latest 1 minute, 1 hour, or 1 day regularly. This requires the time-series database to be able to quickly write and read the latest data. [0004] A typical idea is to persist the time-series data of the last hour or day to the hard disk and at the same time cache it in the memory, so as to realize fast writing and reading of the latest data. [0005] However,...

AI Technical Summary

Problems solved by technology

But it has the following problems: (1) two hours of data are stored in a string structure. When the string capacity is insufficient, a larger space must be applied for and the original data copied to a new space, and the writing performance is relatively low. Poor; (2) When searching for data, it needs to be searched with O(n) time complexity within two hours of string, and the query performance is poor; (3) The index structure is only for time series id, if you need to clear expired data, All time series need to be traversed, and the cleaning efficiency is low; (5) The system has a limitation on data, that is, time series data must be written in order of timestamps, and it does not support out-of-order data writing, which limits the usage scenarios

The problem with this approach is: (1) Using an unbalanced binary sorting tree, the query performance is poor, and in the worst case requires O(n) time complexity; (2) Each node in the binary tree needs to save the left pointer And the right pointer, need to consume a lot of memory, the cost is high

Images