Font Size: a A A

A Design And Implementation Of High Performance Database System For Edge Storage

Posted on:2021-02-02Degree:MasterType:Thesis
Country:ChinaCandidate:B Y SunFull Text:PDF
GTID:2428330620480934Subject:Engineering
Abstract/Summary:PDF Full Text Request
In the past few years,the number of sensors is rapidly increasing on the edge of the net-work.Io T(Internet of things)devices play not only data producers but also data consumers.The traditional cloud computing model adopts centralized management,which has a great possibil-ity of network transmission delay and network fluctuation.It is difficult for cloud computing to meet real-time requirements of applications on the edge.Therefore,edge computing becomes popular as it processes data on the edge directly.In order to store and manage data on the edge,a new way for storage has emerged: edge storage.Edge storage is different from traditional cloud storage.Fist,the mainstream storage medium of Edge storage is different from cloud storage.Flash memory is the mainstream storage medium on the edge,while HDD(hard-disk driver)is the mainstream storage medium of traditional cloud storage.Flash memory can wear out through repeated writes while large amount of data is written on the edge per day.Moreover,the frequency of writes is much higher than reads on the edge and edge storage requires high write performance.Thus,in this paper,Sardine Core is presented for flash memory based on LevelDB with a structure of key-value separation.In Sardine Core,only keys and addresses of values are saved in LSM-tree,which reduces unnecessary data movement in external storage.Therefore,the write amplification of Sardine Core is lowed drastically.Then,the paper implements garbage collection(GC)by divided-and-conquer method,which low the GC burden efficiently.Finally,we implement distributed database SardineDB based on Sardine Core.SardineDB integrates many functions such as decentralized data synchronization,cloud-edge data synchronization,time series service and RESP service.Hence,SardineDB is generic and easily extensible,which can meet the requirement of most Io T applications and sensor data.From evaluation results,the write performance and random read performance of Sardine Core is higher than LevelDB and the SardineDB has great advanes in write performance and random read performance compared with existing distributed databases on the edge.As a result,SardineDB is very suitable for edge,which has a high performance and can meet requirements of devices and data on the edge.
Keywords/Search Tags:IoT, edge computing, LevelDB, sensor, flash memory
PDF Full Text Request
Related items