Font Size: a A A

Research And Implementation Of Interface Transplant Technology In Hadoop Resource Management Module

Posted on:2015-11-21Degree:MasterType:Thesis
Country:ChinaCandidate:Y X LiFull Text:PDF
GTID:2348330542452418Subject:Engineering
Abstract/Summary:PDF Full Text Request
Distributed system and big data platform have become an import technology.Building a good robustness and scalability distributed system is an important way to improve the distributed system and big data field.The Hadoop distributed system is widely used and is a strong potential distributed system.This system is widely used in big data and cloud compute.But under this diversification situation of the distributed system,finding a way to improve it performance and supporting a good compatibility are important for Hadoop system.After the upgrading of Hadoop version,Apache gave out a new resource management module called YARN.Based on a similar resource manager module called EGO,IBM want to transplant the API of YARN module to EGO module.But now there are some problem that we can not transplant it directly.The reason is that there are some isomerism between this two modules.YARN module is developed by JAVA and EGO is developed by C.In addition,this two system have different way to define the structural data.So it need to be solved while doing this work.This paper want to discuss it through the aspects below.1.Firstly,this paper needs to analyse these two system module(YARN module and EGO module).In the analysis,this paper mainly focused on the API and the functions of the module.The concrete content of this paper after is based on the analysis of these two module and API.It should find out the principle of Hadoop YARN module and its API.Besides,it also need to transplant Hadoop YARN API to the thrid-party module according to the characteristic of third-party module.2.Based on a detailed analysis,this paper needs to give out this transplantation work of Hadoop YARN API and the reason why we can not do transplantation directly.According to the problem existed,it needs to use some technology to give out the scheme to solve this problem.Through this scheme,it can transplant the Hadoop YARN API to the EGO module and finish this work.3.At last,this paper needs to do the testing work.Through the unit test of API which has been transplanted,the paper should give out the result of the test which has combined the API and EGO module.Finally,it should point out the problem which still existed and give the suggesstion.For the prospect of the future work,although this paper has finished the work of transplantation of YARN API and finished the testing work.But if the developers want to transplant it completely,still need some extra development work.And analysis the problem which may existed.
Keywords/Search Tags:Big data, Distributed system, Hadoop, API transplantation
PDF Full Text Request
Related items