Abstract. Implementing correct and deterministic parallel programs is challenging. Even though concurrency constructs exist in popular pro-gramming languages to facilitate the task of deterministic parallel pro-gramming, they are often too low level, or do not compose well due to underlying

Abstract. Implementing correct and deterministic parallel programs is challenging. Even though concurrency constructs exist in popular programming languages to facilitate the task of deterministic parallel programming, they are often too low level, or do not compose well due to underlying blocking

, for example, a blocking queue implementation, writing a test case to show that it blocks and unblocks appropriately using existing testing frameworks is exceptionally hard. In this paper, we describe the MultithreadedTC framework which allows the construction of deterministic and repeatable unit tests

serious alternative. This work analyzes the implementation of both a client-server and a peer-to-peer networking model for the prototype shooter game Planet π4. Initially, a survey introduces recent academic approaches to peer-to-peer systems specifically designed for games. Of those, one system

Only a handful of fundamental mechanisms for synchronizing the access of concurrent threads to shared memory are widely implemented and used. These include locks, condition variables, semaphores, barriers, and monitors. In this paper, we introduce a new synchronization mechanism—monotonic counters

In presenting this thesis in partial fulfilment of the requirements for a Postgraduate degree

ii iv vTo my parents vi vii Acknowledgements I would first and foremost extend my whole-hearted gratitude to my advisors, An-drea Arpaci-Dusseau and Remzi Arpaci-Dusseau. Andrea and Remzi are the reason that I had the opportunity for this exceptional Ph.D. journey. To this day, I still re-member the moment when they took me as their student and the joy and hope in my heart. Andrea and Remzi have showed me what systems research is like and how much fun and challenging it can be. Before this journey with them, I had always liked and believed in the beauty of mathematics and theory. My initial interest in systems research happened when I took Remzi’s Advanced Operating Systems

Despite being a core networking primitive, collection protocols today often suffer from poor reliability (e.g., 70%) in practice, and heavily used protocols have never been evaluated in terms of communication efficiency. Using detailed experimental studies, we describe three challenges that cause existing collection protocols to have poor reliability and waste energy: inaccuracies in link estimation, link dynamics, and transient loops. In this paper we present Magnet, a robust, efficient, and hardware-independent collection protocol. Magnet uses three novel techniques to address these challenges. Magnet’s link estimator addresses the inaccuracies in link estimation by using feedback from both the data and control planes, using information from multiple layers through narrow, platform-independent interfaces. Second, Magnet addresses link dynamics by using the Trickle algorithm for control traffic, sending few beacons in stable topologies yet quickly adapting to changes. Finally, Magnet addresses transient loops by using data traffic as active topology probes, quickly discovering and fixing routing failures. Magnet runs on six different mote platforms and we have tested it on four testbeds. In most experiments, Magnet achieves 99 % reliability, and in some cases 99.9%. In the most challenging testbed, the state-of-the-art collection protocol today (MultiHopLQI) achieves 70 % reliability: Magnet achieves 97%. Magnet achieves this ten-fold reduction in dropped packets with 25 % fewer transmissions. Magnet works seamlessly on top of existing low-power MAC layers. Together, these results suggest that Magnet can be the robust, efficient collection layer that so many sensor network applications and protocols need. 1

Abstract: Mobile ad hoc networks (MANETs) and wireless sensor networks (WSNs) are two recently-developed technologies that uniquely function without xed infrastructure support, and sense at scales, resolutions, and durations previously not possible. While both offer great potential in many applications, developing software for these types of networks is extremely dffcult, preventing their wide-spread use. Three primary challenges are (1) the high level of dynamics within the network in terms of changing wireless links and node hardware con gurations, (2) the wide variety of hardware present in these networks, and (3) the extremely limited computational and energy resources available. Until now, the burden of handling these issues was put on the software application developer. This dissertation presents three novel programming models and middleware systems that address these challenges: Limone, Agilla, and Servilla. Limone reliably handles high levels of dynamics within MANETs. It does this through lightweight coordination primitives that make minimal assumptions about network connectivity. Agilla enables self-adaptive WSN applications via the integration of mobile agent and tuple space programming models, which is critical given the continuously changing network. It is the rst system to successfully demonstrate the feasibility of using mobile agents and tuple spaces within

was largely used in more deterministic and discrete ways (dependencies, definitions, alerts). Over time time, we believe organizations will leverage data context using both deterministic and probabilistic approaches. DESIGN AND ARCHITECTURE In a decoupled architecture of multiple applications and backend