Computing and Information Systems - Research Publications
Permanent URI for this collection
Search Results
1 - 3 of 3
-
ItemScalable, security-oriented solutions for nanoCMOS electronics(National e-Science Centre, University of Glasgow, 2008)The EPSRC pilot project Meeting the Design Challenges of nanoCMOS Electronics (nanoCMOS – www.nanocmos.ac.uk) has been funded to tackle some of the challenges facing the semiconductor electronics industry caused by the progressive scaling of CMOS transistors. As transistor dimensions are now at the nanometer scale with 40nm MOSFETs already in mass production and sub-10 nm transistors scheduled for production by 2018, the intrinsic parameter fluctuations caused by the inherent discreteness of charge and matter at this atomistic scale are now one of the major challenges that the semiconductor electronics industry needs to address. The variability at the device level affects profoundly the circuit/system design process and hence can be regarded a semiconductor industry-wide problem. Fortunately many of the statistical variability related issues can be understood and forecasted through large scale simulation of ensembles of potentially hundreds of thousands of atomistically varying devices. However, one of the main distinguishing features of NanoCMOS when compared to other high performance computing (HPC) simulation domains is the imperative requirements on fine grained security. The data, the designs and even the simulations themselves all potentially have highly sensitive commercial intellectual property (IP) value associated with them, ranging from the IP of device manufacturers and the design houses through to licenses needed to run simulation and design software. This paper outlines the e-Infrastructure that has been developed within the nanoCMOS project with specific focus upon the security capabilities it supports and how these address the IP protection requirements of the industrial and collaborating partners. Our ultimate goal is to provide an environment that addresses security across the board and scales to meet the HPC and data management requirements of nanoCMOS research.
-
ItemA resource-oriented data management architecture for nanoCMOS electronics(National e-Science Centre, University of Glasgow, 2008)The EPSRC pilot project Meeting the Design Challenges of NanoCMOS Electronics (nanoCMOS) began in October 2006 and is focused upon tackling the decreasing scale of modern semiconductor components. This scaling has direct impact upon the complete circuit and system electronics design process due to the variability in transistor behaviour caused by differences in atomic structure. To address these challenges, the project has focused upon large scale device simulations exploiting a wide variety of computational resources. This paper focuses on the approach adopted for managing the many hundreds of thousands of files being generated that are associated with these simulations. Specific challenges in achieving this are related to the fine grained security demanded in protecting intellectual property of data and metadata, and the seamless linkage of metadata associated with services. We believe that this data architecture has widespread applicability to many research areas.
-
ItemA functional and performance-oriented comparison of AFS and SRB(National e-Science Centre, University of Glasgow, 2008)In this paper the Storage Resource Broker (SRB) [1] and the Andrew File System (AFS) [2] are compared and contrasted in terms of their use as a data storage for Grid based systems. Both the SRB and AFS are distributed file systems capable of storing data produced and consumed by applications and Grid services. The SRB is well established and deployed by many projects globally include widespread deployment across the UK e-Science National Grid Service. AFS is also a distributed file system initially developed as part of the Andrew Project by Carnegie Mellon University and IBM in 1983. AFS is currently being used by the nanoCMOS project [3, 4] for providing secure access to data and applications through usage of Kerberos tokens. While both SRB and AFS can be used as data back ends they have specific properties that make them better suited for different types of usage. In some cases this may be a critical issue where some feature is required but not available in one or both of the candidates. This paper provides a description of these differences, focusing in particular on the security models, their relative performance with a range of file sizes and their general usability. Based on this analysis we present the key criteria that projects should consider for adoption of file based data management.