基於DC/OS之微服務架構雲端運算平台實作

淡江大學機構典藏 > 工學院 > 資訊工程學系暨研究所 > 學位論文 > Item 987654321/114724

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/114724

Title:	基於DC/OS之微服務架構雲端運算平台實作
Other Titles:	Implementation of microservices for data analytics over DC/OS
Authors:	徐銘志;Hsu, Ming-Chih
Contributors:	淡江大學資訊工程學系碩士班林其誼;Lin, Chi-Yi
Keywords:	Cloud Computing;DC/OS;Kafka;Microservice;Spark;雲端運算;微服物架構
Date:	2017
Issue Date:	2018-08-03 15:01:55 (UTC+8)
Abstract:	隨著雲端計算服務的進步和服務的增長, 微服務架構受到了人們的關注。這一個概念的提出已經有很長一段時間了，但在這幾年卻開始頻繁的出現。微服務架構是一種特定的軟體設計方式-將大型軟體拆分為多個獨立可部署服務組合而成的套件方案。在這個架構出現之前，一個通用的軟體設計模式是使用整體化結構。在這個模式下，應用程式在開發、測試、部署階段都是作為一個整體的存在，這種模式導致擴充應用充滿了挑戰。在這個研究中，我們會在實驗性計算機叢集中，設計一個基於微服務架構資料分析平台，在此我們會使用DC/OS(Data Center Operating System)，一個以Apache Mesos為核心的作業系統，去管理計算機叢集，可以用來管理叢集資源和調度。將這個平台部署在DC/OS可以讓它具有Mesos的高容錯性以及資源共享，也可以輕易地去擴充此平台的功能，讓多種不同類型的服務接合在一起。我們利用了現有的伺服器構建了雲端叢集，在DC/OS上實作資料分析平台，在這個平台上我們主要使用Node.JS、Kafka、Spark來作為我們資料的分析工具，我們的服務會透過DC/OS部屬，而不是透過手動部署。此外，我們使用了Dcard作為我們的資料來源，在即時處理流程我們提供了一個視覺化的介面，去顯示熱門Hashtag，在離線處理流程，透過網頁伺服器，我們可以從資料庫以API的形式，獲取不同類型的資料，像是發文的男女比例或是點讚或評論前十名的文章，展示了這個平台擁有處理即時資料和離線資料的能力，去顯示我們分析完的結果。 With recent advancement and increasing popularity of cloud computing services, the microservice architecture has been gaining more attention in the software development industry. Although the concept of microservices has existed for a long time, its style of architecturing appears more frequently in many software projects until recently. The idea of the microservice architecture is to use a collection of loosely coupled services to compose a large-scale software application. In traditional monolithic architecture, by contrast, every piece of code is put together, and the application is developed, tested, and deployed as a single application. Obviously, it is challenging for the traditional architecture to scale properly. In this research, we implement a data analytics platform based on the microservice architecture over our experimental computer cluster. Specifically, we use DC/OS (Data Center Operating System) to manage the cluster, which extends the functionalities of Apache Mesos to achieve resource management and dispatching. On top of DC/OS, our data analytics platform has the advantages of high fault-tolerance, sharing of system resources, ease of scalability, and well interoperability between different services. Our data analytics platform is built by composing many open-source software such as Node.JS, Apache Kafka, and Apache Spark. The data analytics services will be deployed by DC/OS automatically, without the hassle of error-prone manual deployment. Our data source is Dcard, which is the largest anonymous online community in Taiwan. On streaming processing, our platform offers a visual interface to show the hottest Hashtag. On batch processing, our platform is able to show the statistics such as the ratio of male/female posters, and the top-10 liked or commented posts. The experimental results show that our data analytics platform can do streaming processing and batch processing successfully and reveal useful analytical results.
Appears in Collections:	[資訊工程學系暨研究所] 學位論文

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	173	View/Open

Loading...