分布式服務與服務集群技術研究
發(fā)布時間：2018-11-18 14:54
【摘要】：隨著互聯(lián)網(wǎng)技術的進一步的發(fā)展與深入,越來越多的企業(yè)應用需要應對百萬級的高并發(fā)用戶訪問、低延時及7×24×365高可用性等空前的需求難題,應用系統(tǒng)的業(yè)務訪問量與日俱增,對應用服務器和數(shù)據(jù)庫的要求越來越高。在Web站點以及云平臺中存在大量半結構化數(shù)據(jù),傳統(tǒng)的關系型數(shù)據(jù)庫難以適應半結構化的數(shù)據(jù)存儲訪問模式,例如用戶Session信息。在面臨大規(guī)模數(shù)據(jù)訪問時,RAM與ROM在讀寫速度上相差數(shù)個數(shù)量級,磁盤的IO操作會堆積RAM中的數(shù)據(jù)讀寫請求,造成性能瓶頸,從而導致越來越高的訪問響應延遲。分布式緩存使用系統(tǒng)的高速內(nèi)存(RAM)作為數(shù)據(jù)對象的存儲介質,并采用鍵值對的形式存儲數(shù)據(jù)。這樣就可以在價格低廉、標準化的硬件和軟件平臺上的擴展系統(tǒng)的內(nèi)存,專門用于緩存海量的數(shù)據(jù),為大容量、業(yè)務關鍵型的事務處理提供業(yè)務支撐。隨著業(yè)務需求的進一步發(fā)展,應用規(guī)模在不斷的擴大,巨無霸應用也就越來越多,傳統(tǒng)應用的垂直架構,已經(jīng)無法應對各種復雜業(yè)務帶來的各種挑戰(zhàn)。只有進行業(yè)務梳理,將公共能力抽取成基礎服務或公共服務,對復雜應用進行模塊化拆分和服務化治理,降低服務消費者與服務提供者之間的耦合度。對采用了傳統(tǒng)的垂直架構的系統(tǒng)應用進行服務治理化改造,將應用的各個部分的業(yè)務進行服務抽取和分離,形成一個個相對獨立的服務節(jié)點,其中的每個服務節(jié)點只處理一件或多件相關的同類型的業(yè)務,如單點登陸服務器,只處理用戶登陸相關的業(yè)務。應對百萬級的并發(fā)訪問、每秒數(shù)千的并發(fā)事務處理時,任何單個服務節(jié)點都不能做到及時響應。這個時候,考慮的不是增加單個服務器的性能,而是需要對高訪問量、高并發(fā)量的服務節(jié)點架設一個或多個服務器集群分擔業(yè)務壓力,形成多臺服務器提供相同的服務,并相互之間相互獨立的服務器集群,最終實現(xiàn)整個應用的垂直架構平滑地向分布式服務框架演變。本論文選題來自于本人曾經(jīng)實習的亞信科技有限公司的服務治理項目組,靈感來源于其對北京移動CRM系統(tǒng)服務化改造的北京移動能力開放平臺項目,能力開放平臺通過將業(yè)務支撐系統(tǒng)業(yè)務能力的逐步有序開放與管控,實現(xiàn)內(nèi)部業(yè)務能力對內(nèi)、外部渠道的共享以及開發(fā)者和運營者的利益雙贏,使得電信業(yè)務開發(fā)變得更為容易和高效,促進移動互聯(lián)網(wǎng)業(yè)務的發(fā)展和繁榮,支持業(yè)務能力持續(xù)創(chuàng)新。本文旨在研究和探討,如何實現(xiàn)通過服務治理的方式將傳統(tǒng)應用的垂直架構方案演變成分布式服務架構方案,以達到解耦服務提供方和服務使用方之間的耦合關系的目的,并通過分布式緩存、分布式服務、服務器集群、消息中間件、服務注冊中心、數(shù)據(jù)庫分庫分表、負載均衡、系統(tǒng)業(yè)務分拆等技術手段應對高訪問量、高并發(fā)量、高可用等方面的應用需求。
[Abstract]:With the further development and deepening of Internet technology, more and more enterprise applications need to deal with the unprecedented difficulties of high concurrent user access, low latency and 7 脳 24 脳 365 high availability, and the number of business visits to the application system is increasing day by day. Application servers and databases are increasingly demanding. There is a large amount of semi-structured data in Web sites and cloud platforms, and traditional relational databases are difficult to adapt to semi-structured data storage access patterns, such as user Session information. When faced with large-scale data access, RAM and ROM differ several orders of magnitude in read and write speed. The IO operation of disk will pile up data read and write requests in RAM, resulting in performance bottleneck, which leads to higher and higher access response delay. The distributed cache uses the high speed memory (RAM) of the system as the storage medium of the data object and stores the data in the form of key-value pairs. In this way, we can extend the memory of the system on the cheap, standardized hardware and software platform, specially used to cache the massive data, and provide the service support for the transaction processing of the large capacity and business critical type. With the further development of business requirements and the continuous expansion of application scale, there are more and more Big Mac applications. The vertical structure of traditional applications can no longer cope with various challenges brought by complex business. Only by combing the business, extracting the public capacity into basic services or public services, the complex applications are divided into modular and service-oriented governance, so as to reduce the coupling degree between service consumers and service providers. The system application with traditional vertical architecture is transformed into service governance, and the services of each part of the application are extracted and separated to form a relatively independent service node. Each service node only handles one or more related services, such as single-point login server, and only deals with user logon related services. For millions of concurrent access and thousands of concurrent transactions per second, no single service node can respond in a timely manner. At this time, the consideration is not to increase the performance of a single server, but to set up one or more server clusters with high traffic and high concurrency to share the business pressure, resulting in multiple servers providing the same service. Finally, the vertical architecture of the whole application evolves smoothly to the distributed service framework. The topic of this thesis comes from the service management project team of Asense Technology Co., Ltd., which I once worked as an intern. The inspiration comes from the Beijing Mobile open platform project for the service transformation of Beijing Mobile CRM system. The open platform realizes the sharing of internal and external channels and the interests of developers and operators through the gradual and orderly opening and control of the business capabilities of the business support system. It makes telecom service development easier and more efficient, promotes the development and prosperity of mobile Internet service, and supports continuous innovation of service capability. The purpose of this paper is to study and discuss how to transform the vertical architecture scheme of traditional application into a distributed service architecture scheme through service governance, in order to decouple the coupling relationship between service provider and service consumer. And through distributed cache, distributed services, server cluster, message middleware, service registry, database table, load balancing, system business separation and other technical means to deal with high traffic, high concurrency, etc. High availability and other application requirements.
【學位授予單位】：成都理工大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP393.092
本文編號：2340349
[Abstract]:With the further development and deepening of Internet technology, more and more enterprise applications need to deal with the unprecedented difficulties of high concurrent user access, low latency and 7 脳 24 脳 365 high availability, and the number of business visits to the application system is increasing day by day. Application servers and databases are increasingly demanding. There is a large amount of semi-structured data in Web sites and cloud platforms, and traditional relational databases are difficult to adapt to semi-structured data storage access patterns, such as user Session information. When faced with large-scale data access, RAM and ROM differ several orders of magnitude in read and write speed. The IO operation of disk will pile up data read and write requests in RAM, resulting in performance bottleneck, which leads to higher and higher access response delay. The distributed cache uses the high speed memory (RAM) of the system as the storage medium of the data object and stores the data in the form of key-value pairs. In this way, we can extend the memory of the system on the cheap, standardized hardware and software platform, specially used to cache the massive data, and provide the service support for the transaction processing of the large capacity and business critical type. With the further development of business requirements and the continuous expansion of application scale, there are more and more Big Mac applications. The vertical structure of traditional applications can no longer cope with various challenges brought by complex business. Only by combing the business, extracting the public capacity into basic services or public services, the complex applications are divided into modular and service-oriented governance, so as to reduce the coupling degree between service consumers and service providers. The system application with traditional vertical architecture is transformed into service governance, and the services of each part of the application are extracted and separated to form a relatively independent service node. Each service node only handles one or more related services, such as single-point login server, and only deals with user logon related services. For millions of concurrent access and thousands of concurrent transactions per second, no single service node can respond in a timely manner. At this time, the consideration is not to increase the performance of a single server, but to set up one or more server clusters with high traffic and high concurrency to share the business pressure, resulting in multiple servers providing the same service. Finally, the vertical architecture of the whole application evolves smoothly to the distributed service framework. The topic of this thesis comes from the service management project team of Asense Technology Co., Ltd., which I once worked as an intern. The inspiration comes from the Beijing Mobile open platform project for the service transformation of Beijing Mobile CRM system. The open platform realizes the sharing of internal and external channels and the interests of developers and operators through the gradual and orderly opening and control of the business capabilities of the business support system. It makes telecom service development easier and more efficient, promotes the development and prosperity of mobile Internet service, and supports continuous innovation of service capability. The purpose of this paper is to study and discuss how to transform the vertical architecture scheme of traditional application into a distributed service architecture scheme through service governance, in order to decouple the coupling relationship between service provider and service consumer. And through distributed cache, distributed services, server cluster, message middleware, service registry, database table, load balancing, system business separation and other technical means to deal with high traffic, high concurrency, etc. High availability and other application requirements.
【學位授予單位】：成都理工大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP393.092
【參考文獻】
相關博士學位論文 前2條
1 王紅斌;Web服務器集群系統(tǒng)的自適應負載均衡調度策略研究[D];吉林大學;2013年
2 盧本捷;分布式消息隊列的理論、實現(xiàn)與應用[D];華中科技大學;2004年
相關碩士學位論文 前10條
1 張羽;一種分布式服務治理框架的設計與實現(xiàn)[D];北京交通大學;2016年
2 劉先紡;基于Dubbo的網(wǎng)上訂餐系統(tǒng)[D];華南理工大學;2016年
3 張永祥;電子商務系統(tǒng)平臺模型的研究與實現(xiàn)[D];電子科技大學;2016年
4 馮博;基于PaaS的電信虛擬運營商系統(tǒng)設計和實現(xiàn)[D];中國科學院大學(工程管理與信息技術學院);2015年
5 王利萍;基于Nginx服務器集群負載均衡技術的研究與改進[D];山東大學;2015年
6 侯yN寒;電信能力開放平臺中鑒權框架的設計與實現(xiàn)[D];北京郵電大學;2015年
7 崔解賓;分布式內(nèi)存緩存技術在數(shù)據(jù)處理平臺中的研究與應用[D];北京郵電大學;2015年
8 閆明;高可用可擴展集群化Redis設計與實現(xiàn)[D];西安電子科技大學;2014年
9 沈偉;云計算平臺下分布式緩存系統(tǒng)的性能優(yōu)化研究[D];西安電子科技大學;2014年
10 吳偉華;基于分布式緩存的內(nèi)容分發(fā)網(wǎng)絡[D];西安電子科技大學;2014年
，本文編號：2340349
本文鏈接：http://sikaile.net/guanlilunwen/ydhl/2340349.html
最近更新
教材專著
