
Context-Aware Task Scheduling in Distributed Computing Systems Inauguraldissertation zur Erlangung des akademischen Grades eines Doktors der Wirtschaftswissenschaften der Universit¨atMannheim vorgelegt von Janick Edinger aus Mannheim ii Prodekan: Prof. Dr. Moritz Fleischmann Referent: Prof. Dr. Christian Becker Korreferent: Prof. Dr. Matti Rossi Tag der m¨undlichen Pr¨ufung:29. April 2019 Pr¨ufungsausschuss: Prof. Dr. Christian Becker (Vorsitzender) Prof. Dr. Hartmut H¨ohle iii For my family, who taught me trust. For Laura, who taught me love. For Lion, who taught me priorities. For Dominik, who taught me that 4a.m. is the perfect time for new research ideas. Abstract These days, the popularity of technologies such as machine learning, augmented reality, and big data analytics is growing dramatically. This leads to a higher demand of computational power not only for IT professionals but also for ordinary device users who benefit from new applications. At the same time, the computa- tional performance of end-user devices increases to meet the demands of these resource-hungry applications. As a result, there is a coexistence of a huge demand of computational power on the one side and a large pool of computational resources on the other side. Bringing these two sides together is the idea of computational resource sharing systems which allow applications to forward computationally intensive workload to remote resources. This technique is often used in cloud computing where customers can rent computational power. However, we argue that not only cloud resources can be used as offloading targets. Rather, idle CPU cycles from end-user administered devices at the edge of the network can be spon- taneously leveraged as well. Edge devices, however, are not only heterogeneous in their hardware and software capabilities, they also do not provide any guarantees in terms of reliability or performance. Does it mean that either the applications that require further guarantees or the unpredictable resources need to be excluded from such a sharing system? In this thesis, we propose a solution to this problem by introducing the Tasklet system, our approach for a computational resource sharing system. The Tasklet system supports computation offloading to arbitrary types of devices, including stable cloud instances as well as unpredictable end-user owned edge resources. Therefore, the Tasklet system is structured into multiple layers. The lowest layer is a best-effort resource sharing system which provides lightweight task scheduling and execution. Here, best-effort means that in case of a failure, the task execution is dropped and that tasks are allocated to resources randomly. To provide execution guarantees such as a reliable or timely execution, we add a Quality of Computation (QoC) layer on top of the best-effort execution layer. The QoC layer enforces the guarantees for applications by using a context-aware task scheduler which monitors the available resources in the computing environment and performs the matchmaking between resources and tasks based on the current state of the system. As edge resources are controlled by individuals, we consider the fact that these users need to be able to decide with whom they want to share their resources and for which price. Thus, we add a social layer on top of the system that allows users to establish friendship connections which can then be leveraged for social-aware task allocation and accounting of shared computation. v Acknowledgments This thesis would not have been possible without the support of many people. Here, I would like to take the chance and thank all these people who have supported me throughout the past six years in so many different ways. First, I would like to thank Prof. Dr. Christian Becker for the guidance and support during my time at the chair of Information Systems II. Christian, listing all the anecdotes from the past years here, would easily fill another 200 pages. Instead, I simply want to say thank you for being a teacher when I needed to catch up in computer science, for being a supervisor when I needed advice for a paper, for being a counselor when I needed to make work or life decisions, and for being a friend when we traveled Boston, Budapest, Hong Kong, Hawaii, Athens, and Kyoto. You always managed to find the fine line between giving me guidance when necessary and providing me the opportunity to follow my own ideas. I am thankful for all of this. I would like to thank Prof. Dr. Hartmut H¨ohlefor accepting me in his team and for introducing me to enterprise systems research. With every conversation we have, I understand a little more about theoretical contributions. I am looking forward to all the projects that we have already planned together. I would like to thank Prof. Dr. Matti Rossi, who immediately agreed to be a reviewer for this thesis. It would be an honor for me to work together with you in future research projects. During my time as a (PhD) student there were multiple people who I consider as mentors as they tought me so much about life in general, and in particular, research and teaching. First of all, there is Prof. Dr. Gregor Schiele, who already believed in me when I was still struggling to write my first lines of code in Java as a student. Gregor, thank you for always taking time when I needed your help or advice. I am grateful for the support of Dr. Justin Mazzola Paluska who mentored us in the very beginning of the Tasklet project. Our meetings at MIT are some of the most memorable experiences during my time as a PhD student. Dr. Sebastian VanSyckel, or Seb for short, thank you for taking the lead in my first Tasklet paper and for supporting so many of our papers after this. Finally, thank you Philipp (Schaber) for your endless amount of ideas and your critical assessment of my papers. Especially, when I asked for it. Thank you Dr. Miriam Spering, Prof. Dr. Dinesh Pai, and Prof. Dr. Jiannong Cao for hosting me during my stays at the University of British Columbia and vii viii the Hong Kong Polytechnic University. Miriam, thank you for introducing me to the fascinating world of visual sciences and for keeping up this collaboration and friendship over so many years until today. Many thanks to the best team I can imagine that provided me with a pleasant and inspiring working environment. Especially in the most stressful and demanding times in which we spent day and night working towards deadlines, the level of helpfulness and team spirit was beyond words. Thank you to Dr. Patricia Arias-Cabarcos, Martin Breitbach, Kerstin Goldner, Melanie Heck, Benedikt Kirpes, Sonja Klingert, Dr. Christian Krupitzer, Markus Latz, Jens Naber, Yugo Nakamura, Martin Pfannem¨uller,Dr. Vaskar Raychoudhury, Dr. Felix Maximilian Roth, Dr. Dominik Sch¨afer,and Anton Wachner. In particular, I want to thank Dominik, for all the time that we have spent working together starting from the very early days as student assistants. Your dedication to the project in endless night shifts, your optimism in situations that seemed hopeless (`What should go wrong?'), and your friendship made you the ideal project partner. I will miss our discussions. Pitzi, thank you for all the years of managing us and for joining multiple night shifts to get the most out of every paper. I admire your discipline and loyalty. Thank you Max and Jens for lightening my workload especially during the last year which allowed me to work on this thesis. Pfanni, thank you for always being there when help is needed and for your continuous improvements for our team (Git server, exam templates, registration tool, CT2 lectures, ...). To the Breitis: Thank you for all your support over the years as student assistant, co-author, and colleague and for continuing the Tasklet project in the future. Thank you, Anton, as working with you in the EyeT Lab project is always exciting and a great experience. Also, many thanks to you, Kerstin, for all your support in administrative tasks throughout the years. Thank you to the team of the chair of Enterprise Systems for the warm welcome and for sharing your research ideas with me. This includes Karoline Glaser, Brigitte Haber, Tobias Nisius, Florian Pethig, and Fareed Zandkarimi. Thank you to my co-authors from other groups and universities. The joint work across the boundaries of departments and countries makes research much more engaging. Thank you Prof. Dr. Teruo Higashino, Prof. Dr. Andy Sch¨urr, Prof. Dr. Heiner Stuckenschmidt, Dr. Hirozumi Yamaguchi, Tim Bauer, Bastian Biemer, Sunyanan Choochotkaew, Alexandra Hofmann, Robin Klose, Roland Kluge, Manisha Luthra, Timo Sztyler, and Markus Weckesser. This work was supported by the German Research Foundation (DFG) under grant BE 2498=10−1 `Tasklets: Ein Ansatz f¨urBest-Effort Computing'. Thanks to the students who supported the Tasklet project within their theses and beyond. They are in particular: Erik Baumert, Tobias Borlinghaus, Tianhong Chen, Simona Doneva, Lukas Draheim, Anna Florentine Eitelbach, Daniel Flachs, Niklas Fr¨uhauf,Philipp Kaiser, Martin Schaumann, Nadja Seemann, and Aleksander Stelmaszczyk. ix I would like to thank my family who has provided me with unconditional support in every stage of my life. At all times, you never hesitated to help out even though you had to cancel or change your own plans. To my parents, thank you for always being there for me and for providing me with everything that I needed to pursue my own way. To my sister Nina, thank you for teaching me how to read when we were children. This has turned out to be very useful, indeed. Thanks also to Helga, Wolfgang, and Lisa for babysitting, cooking, and housekeeping so many times during the final stage of this thesis.
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages290 Page
-
File Size-