Design And Implementation Of Time Series Data Cache Based On Memcached

In the monitoring system of clusters and Io T devices,a large amount of high-dimensional time series data is generated.In order to efficiently store and analyze these massive amounts of data in real time,time series data management systems have been developed.As an open source KV cache system based entirely on memory,memcached has been widely concerned by the outside world.Memcached completely stores data in memory,and has an efficient and concise memory management mechanism,which is very suitable for storage and analysis of time series data with high real-time requirements.Moreover,memcached is currently not optimized for efficient storage of time series data,so the re-search on time series data caching based on memcached is an important research topic.Thesis studies and implements a cache system MEMTS based on memcached time series data,which is used for efficient storage and low latency query and analysis of recent time series data.In this paper,the main research contents are:(1)A memory-based time series data storage engine is designed based on mem-cached.According to the characteristics of time series data,a double-layer memory struc-ture is designed to support efficient writing and low-latency reading? in order to reduce memory usage,Gorilla compression algorithm and Slicing Bitmap are used? For efficient aggregation query,a PISA index is used for pre-aggregation? in order to ensure sufficient memory,an improved periodic deletion memory management strategy is designed.(2)The cache management system of MEMTS time series data is realized.On the basis of native memcached,the time series data cache management system designed in this paper is implemented,including efficient memory segmentation structure,Gorilla data compression algorithm,PISA pre-aggregation index,Slicing Bitmap index,memory management strategy,etc.(3)Extensive experiments are carried out to test the function and performance of MEMTS.This article first tests the basic functions of MEMTS,and then selects two open source memory time series databases,Beringei and Prometheus,for performance testing.The results show that MEMTS functions normally and performs well in query efficiency and memory footprint.