Large-scale graph-structured datasets are growing at an increasing rate. Social network graphs are an example of these datasets. Processing large-scale graphstructured datasets are central to many applications ranging from telecommunication to biology and has led to the development of many parallel graph algorithms. Performance of parallel graph algorithms largely depends on how the underlying graph is partitioned. In this work, we focus on studying streaming vertex-cut graph partitioning algorithms where partitioners receive a graph as a stream of vertices and edges and assign partitions to them on their arrival once and for all. Some of these algorithms maintain a state during partitioning. In some cases, the size of the state is so huge that it cannot be kept in a single machine memory. In many real world scenarios, several instances of a streaming graph partitioning algorithm are run simultaneously to improve the system throughput. However, running several instances of a partitioner drops the partitioning quality considerably due to the incomplete information of partitioners. Even frequently sharing states and its combination with buffering mechanisms does not completely solves the problem because of the heavy communication overhead produced by partitioners. In this thesis, we propose an algorithm which tackles the problem of low scalability and performance of existing streaming graph partitioning algorithms by providing an efficient way of sharing states and its combination with windowing mechanism. We compare state-of-the-art streaming graph partitioning algorithms with our proposed solution concerning performance and efficiency. Our solution combines a batch processing method with a shared-state mechanism to achieve both an outstanding performance and a high partitioning quality. Shared state mechanism is used for sharing states of partitioners. We provide a robust implementation of our method in a PowerGraph framework. Furthermore, we empirically evaluate the impact of partitioning quality on how graph algorithms perform in a real cloud environment. The results show that our proposed method outperforms other algorithms in terms of partitioning quality and resource consumption and improves partitioning time considerably. On average our method improves partitioning time by 23%, decreases communication load by 15% and increase memory consumption by only 5% compared to the state-of-the-art streaming graph partitioning.

The quantity of rich, semi-structured data generated by sensor networks, scientific simulation, business activity, and the Internet grows daily. The objective of this research is to investigate architectural requirements for emerging applications in massive graph analysis. Using emerging hybrid systems, we will map applications to architectures and close the loop between software and hardware design in this application space. Parallel algorithms and specialized machine architectures are necessary to handle the immense size and rate of change of today's graph data. To highlight the impact of this work, we describe a number of relevant application areas ranging from biology to business and cybersecurity. With several proposed architectures for massively parallel graph analysis, we investigate the interplay of hardware, algorithm, data, and programming model through real-world experiments and simulations. We demonstrate techniques for obtaining parallel scaling on multithreaded systems using graph algorithms that are orders of magnitude faster and larger than the state of the art. The outcome of this work is a proposed hybrid architecture for massive-scale analytics that leverages key aspects of data-parallel and highly multithreaded systems. In simulations, the hybrid systems incorporating a mix of multithreaded, shared memory systems and solid state disks performed up to twice as fast as either homogeneous system alone on graphs with as many as 18 trillion edges.