No abstract

Replication has recently gained attention in the context of fault tolerance for large scale MPI HPC applications. Existing implementations try to cover all MPI codes and to be independent from the underlying library. In this paper, we evaluate the advantages of adopting a different approach. First, we try to take advantage of a communication property common to many MPI HPC application, namely senddeterminism. Second, we choose to implement replication inside the MPI library. The main advantage of our approach is simplicity. While being only a small patch to the Open MPI library, our solution called SDR-MPI supports most main features of the MPI standard including all collectives and group operations. SDR-MPI additionally achieves good performance: Experiments run with HPC benchmarks and applications show that its overhead remains below 5%.

Before deploying their infrastructure (resources, data, communications, ...) on a Cloud computing platform, companies want to be sure that it will be properly secured. At deployment time, the company provides a security policy describing its security requirements through a set of properties. Once its infrastructure deployed, the company want to be assured that this policy is applied and enforced. But describing and enforcing security properties and getting strong evidences of it is a complex task.To address this issue, in [1], we have proposed a language that can be used to express both security and assurance properties on distributed resources. Then, we have shown how these global properties can be cut into a set of properties to be enforced locally. In this paper, we show how these local properties can be used to automatically configure security mechanisms. Our language is context-based which allows it to be easily adapted to any resource naming systems e.g., Linux and Android (with SELinux) or PostgreSQL. Moreover, by abstracting low-level functionalities (e.g., deny write to a file) through capabilities, our language remains independent from the security mechanisms. These capabilities can then be combined into security and assurance properties in order to provide high-level functionalities, such as confidentiality or integrity. Furthermore, we propose a global architecture that receives these properties and automatically configures the security and assurance mechanisms accordingly. Finally, we express the security and assurance policies of an industrial environment for a commercialized product and show how its security is enforced.