|Line: 1 to 1|
|h2 Project Description|
In this project, funded by the KAREN Capability Fund, the Network Design and Simulation Group will apply the optimization techniques they have developed for optimal Storage Area Network design to Data GRID . This project is an exemplar for use of Computational GRID to conduct computer-intensive simulation and optimization research, establishing a decision framework for effective performance of Data GRID.
A subsequent project could extend this approach into modeling and optimization of Computational GRID and Access GRID, with a view to ultimately developing a best practice capacity planning framework for the KAREN NREN infrastructure and BeSTGRID Data GRID services. This would be achieved by working collaboratively with the WAND Network Research Group, specifically their project team working on KAREN network measurement and modeling .
h2 Current Work
[DataGRID Network Simulation]
* [BeSTGRID SAN Meeting Minutes 20071130]
* [BeSTGRID SAN Meeting Minutes 20071105]
h2 Project Phases
The project has a 3-phased approach to accurately model Data GRID and plan for its future reprovisioning.
h6 Phase 1:
* Establishing the Computational GRID modeling environment;
* Building statistical models of initial BeSTGRID Loads and Flows;
* Building a computer simulation of Data GRID to model its performance under initial demand;
h6 Phase 2:
* Determining probable scenarios (conservative/expected/high) of user adoption within New Zealand to model Data GRID’s traffic levels into the future;
h6 Phase 3:
* Using mixed-integer programming together with computer simulation to develop a design for reprovisioning Data GRID that will support these future requirements most “effectively”. The approaches are either:
# provide a scalable network with multiple disjoint flow-paths and fixed over-subscription ratio at lowest cost; or
# provide a scalable network with multiple disjoint flow-paths and fixed cost with minimum over-subscription ratio.
h2 Project Timeline
The first phase of this project will be completed by July 2008. It will involve:
* the acquisition of knowledge (and [data|SAN Traffic Logging]) regarding the existing Data GRID;
* the development of statistical models to describe the current state of Data GRID;
* the building of a computer model (using OpNet) to simulate Data GRID;
* calibration and validation of the model;
* reporting on Data GRID’s current performance.
The second phase of this project will be completed by August 2008. It will involve:
* the development of probable scenarios for KAREN’S adoption by New Zealand researchers: under conservative/expected/high adoption rates, 1/2/3 years into the future
* the conversion of these scenarios into expected traffic levels to drive our simulation models
The third and final phase will be completed by June 2009. It will involve:
* the use of optimisation models to design reprovisioning plans for Data GRID under the traffic flows determined in phase 2
* the extension of the existing computer simulation to model these new designs
* the testing of these reprovisioned Data GRIDs via simulation under the traffic rates determined in phase 2.
h2 Alignment with Goals of the KAREN Capability Fund
The project is in line with the exemplar project and infrastructure activity outlined in the KAREN Working Group and Development Fund framework. It will play a crucial part in the achievement of all of the Fund’s goals:
# Establish awareness – this project will demonstrate the viability of using KAREN for sophisticated optimization and computer-intensive simulation modelling;
# Enable effective use of the advanced network – by optimising Data GRID’s performance it will significantly improve the effectiveness of KAREN’s use;
# Promote the use of KAREN – it will promote KAREN’s use in computer simulation and optimization research;
# Create a community ethos of sharing knowledge – optimising Data GRID’s performance will play a significant role in avoiding user dissatisfaction, which is one potential barrier to the creation of a community ethos of knowledge sharing.
This proposal requires the analysis of Data GRID’s network data , as well as the use of computer-intensive simulation and optimization models . We will store this data and run these simulation models on KAREN. The resulting computer simulation and optimization models will be made available to KAREN for use into the future for ongoing decision analysis and planning.
h2 Project Investigators
__Cameron Walker__ is a Senior Lecturer and member of the Operations Research Group in the Department of Engineering Science. His research interests are Network Optimisation, Simulation, and Statistical modeling. He has been involved in research on Optimal Design of Storage Area Networks (SANs) since 2002.
__Michael O’Sullivan__ is a Lecturer and member of the Operations Research Group in the Department of Engineering Science. He began his research into the design of SANs in 1998 while working on his PhD at Stanford University. He worked as a consultant to Hewlett-Packard Laboratories and was involved in numerous patents for automatic SAN design using heuristic algorithms. Upon accepting a position in the Department of Engineering Science at the University of Auckland he formed the Network Design and Simulation Group with Cameron Walker. They have maintained their relationship with HP Labs, Palo Alto, and also collaborate with the Storage Systems Research Center at UC Santa Cruz.
__Ilze Ziedins__’ main research interest is the modeling and analysis of stochastic networks, particularly telecommunications and computer networks. Since completing her PhD at Cambridge, she has worked at Heriot-Watt and Auckland Universities. She has done major consultancy work for Telecom and Bell Labs in New Jersey, with patents arising from the latter. She has been working with the WAND group for several years on statistical models of traffic in networks.
__Nevil Brownlee__ was Architect of the University of Auckland campus network from its beginnings in 1986 until 1998. From 1992 he has been active in the IETF (Internet Engineering Task Force), chairing several of its Working Groups. Currently he chairs its IP Flow Export (IPFIX) working group, which is producing a standard method for exporting flow information from network devices. From 2000 to 2003 Nevil spent 50% of his time as a Researcher at CAIDA, the Cooperative Association for Internet Data Analysis at UC San Diego, where he worked on various network measurement projects. His association with CAIDA continues, in the form of a long-term research effort into the stability and security of the global Domain Name System (DNS). Since February 2004 he has been an Associate Professor in Computer Science at the University of Auckland, where he continues his Network Measurement research and teaches Data Communications and Networking.
__Nick Jones begin_of_the_skype_highlighting end_of_the_skype_highlighting__’ interest is in the area of industrial applications of research driven ICT. Nick is R & D Strategist for the Network Design and Simulation Group, Chief Technical Officer for e-learnings (a University of Auckland business at UniServices), and Project Manager within the Centre for Software Innovation working on BeSTGRID and the ICT Academy projects among others. He joined the Network Design and Simulation Group in 2005 and has since been active in developing the group’s industry links and in helping refine a research direction that focuses on industry applications.
-- MichaelOSullivan - 16 Dec 2010 \ No newline at end of file