Thèse de doctorat--Électronique et communications--Paris--ENST, 2006. / Notes bibliogr. Glossaire. Résumé en français et en anglais.

Thèse de doctorat--Électronique et communications--Paris--ENST, 2006. / SIMO = Single input multiple output. Bibliogr. p. 167-187. Résumé en français et en anglais.

Thesis (M.S.)--Boise State University, 2009. / Title from t.p. of PDF file (viewed Mar. 15, 2010). Includes abstract. Includes bibliographical references (leaves 78-82).

Not available / text

To make progress in the face of failures, long-running parallel applications need to save their state, known as a checkpoint. Unfortunately, current checkpointing techniques are becoming untenable on large-scale supercomputers. Many applications checkpoint all processes simultaneously--a technique that is easy to implement but often saturates the network and file system, causing a significant increase in checkpoint overhead. This thesis introduces compiler-assisted staggered checkpointing, where processes checkpoint at different places in the application text, thereby reducing contention for the network and file system. This checkpointing technique is algorithmically challenging since the number of possible solutions is enormous and the number of desirable solutions is small, but we have developed a compiler algorithm that both places staggered checkpoints in an application and ensures that the solution is desirable. This algorithm successfully places staggered checkpoints in parallel applications configured to use tens of thousands of processes. For our benchmarks, this algorithm successfully finds and places useful recovery lines that are up to 37% faster for all configurations than recovery lines where all processes write their data at approximately the same time. We also analyze the success of staggered checkpointing by investigating sets of application and system characteristics for which it reduces network and file system contention. We find that for many configurations, staggered checkpointing reduces both checkpointing time and overall execution time. To perform these analyses, we develop an event-driven simulator for large-scale systems that estimates the behavior of the network, global file system, and local hardware using predictive models. Our simulator allows us to accurately study applications that have thousands of processes; it on average predicts execution times as 83% of their measured value. / text

No description available.

No description available.

No description available.

No description available.

As more applications rely on distributed systems (peer-to-peer services, content distribution networks, cloud services), it becomes necessary to identify hosts that return content to the user with minimal delay. A large scale map of delays would aid in solving this problem. Existing methods, which deploy devices to every region of the Internet or use of a single vantage point have yet to create such a map. While services such as PlanetLab offer a distributed network for measurements, they only cover 0.3% of the Internet. The focus of our research is to increase the speed of the single vantage point approach so that it becomes a feasible solution. We evaluate the feasibility of performing large scale measurements by performing an experiment using more hosts than any previous study. First, an efficient scanning algorithm is developed to perform the measurement scan. We then find that a custom Windows network driver is required to overcome bottlenecks in the operating system. After developing a custom driver, we perform a measurement scan larger than any previous study. Analysis of the results reveals previously unidentified drawbacks to the existing architectures and measurement methodologies. We propose novel meth- ods for increasing the speed of experiments, improving the accuracy of measurement results, and reducing the amount of traffic generated by the scan. Finally, we present architectures for performing an Internet scale measurement scan. We found that with custom drivers, the Windows operating system is a capable platform for performing large scale measurements. Scan results showed that in the eleven years since the original measurement technique was developed, the response patterns it relied upon had changed from what was expected. With our suggested improvements to the measurement algorithm and proposed scanning architectures, it may be possible to perform Internet scale measurement studies in the future.