Master Thesis MSTR-2018-94

BibliographyUl Haq, Hafiz Irfan: Configuration Management Platform for the Automation of the Bosch IoT Cloud.
University of Stuttgart, Faculty of Computer Science, Electrical Engineering, and Information Technology, Master Thesis No. 94 (2018).
59 pages, english.
Abstract

With the advancement of high-speed networks over the last decades, cloud computing technology focus has been shifted towards deploying massive de-centralized datacenters comprises of hundreds and thousands of nodes. This huge network of data centers distributed among many different regions around the world. These distributed cloud platform required several hardware and software components that needed to be appropriately configured. The configuration of components in regions needs to be consistent to ease the operations overhead. However, due to the vast amount of infrastructure components that are known to suffer from hardware and software failures, it is likely that network partitions or node failures occur. In consideration of the CAP theorem, it is hard to manage the cloud’s configuration due to system failures. Even when one or more regions are unavailable due to network failure, cloud operators still need to be able to maintain and manage the cloud configurations of other regions. Inconsistencies of components due to region failures that came back up after a failure leads to a tremendous amount of time and cost to manage and resolve the inconsistency in the distributed cloud data centers. This thesis presents the design and implementation of Global Configuration Management Platform, a highly available and eventually consistent system that automate the process of managing cloud configuration. To achieve this level of availability, GCMP sacrifices consistency under specific failure situations. It makes use of configuration versioning and provides conflict resolution in a manner that provides a novel interface for cloud operators to use. The GCMP server uses two separate databases to store global and local configurational data to resolve availability and configurational inconsistency issues. Moreover, the web portal to configure cloud also distributed over a different region, which gives us the advantage over to configure the system from outside the region even in the presence of network failure within the region. Experimental results show that the GCMP system performs well and makes the configurational data eventually consistent under many failure conditions such as database failure, message broker failure, server or client failures. It can also provide versioning and out-of-order update request from the cloud operators. The scope of our work is to finding the tradeoff between consistency and availability model to configure the Bosch IoT cloud, which is one example of the distributed cloud platform. Similarly, our design model of GCMP can be applied to any other cloud-based infrastructure.

Department(s)University of Stuttgart, Institute of Parallel and Distributed Systems, Distributed Systems
Superviser(s)Rothermel, Prof. Kurt; Bibartiu, Otto; Ottenwälder, Dr. Beate; Grimm, Philipp
Entry dateJune 18, 2019
   Publ. Department   Publ. Institute   Publ. Computer Science