Faculty of Engineering


b-energy-green-storage

Energy efficient data storage systems

The goal of this research project is to integrate power consumption into existing optimisation models for automatic design of clustered data storage systems, leading to reduced energy consumption.

The recent move to cloud computing has seen the demand for computing resources grow rapidly. For example, the explosion of information has seen exponential growth in demand for cloud storage. This has lead to a corresponding increase in the size and number of servers within a cloud infrastructure and a consequent increase in energy required to operate and cool them. Clustered cloud infrastructures are now being used as a low-cost alternative to monolithic systems, and they have the additional potential for considerable power savings. This research investigates incorporation of power consumption into the design of cloud infrastructure. 

Key focus areas/issues


How sustainable is the growth in demand for energy?

Concern for current and future energy levels has lead to new benchmarks for technological price-performance comparisons. The goal of this research is to integrate power consumption into existing optimisation models for automatic design of cloud infrastructure, leading to reduced energy consumption. This involves:

  • Characterising the performance of the components of the cloud in terms of energy consumption under typical traffic loads
  • Analysis of the resulting data and characterisation of the power use profiles of the system components
  • Modelling power consumption in the design processfor cloud infrastructure
  • Validation via implementation of a test cloud
cl-energy-power-testingboard

Current major developments


This year we are completing rigorous testing of a cloud storage infrastructure, as well as experimenting with OpenStack, a new Cloud OS growing in popularity. Our Cloud Technology Lab within the Department of Engineering Science is set up for measuring the power consumption of disks, servers and cloud infrastructure under “typical” loads. Previous researchers in the Cloud Technology Lab include a post-doctoral fellow, DongJin Lee (now working at Samsung Electronics) and three international interns (two from IITs in India and one from the Netherlands) who each visited the lab for 3 months to work on cloud technology research projects.

Key achievements


We have implemented a prototype OpenStack installation and compared its performance with its native storage platform (Swift) and a leading distributed storage platform (Ceph). We have a test installation in the university data centre running Ceph. We have embedded an open-source version of the industry testing standard (SPC-1) within an open source workload generator (fio) to provide a benchmark tool for the research community. We have used this benchmark to characterise the performance of high- and low-end SSD and HDD storage in terms of performance and energy consumption under typical traffic loads.

Key people


Contact


Cameron Walker
Email: cameron.walker@auckland.ac.nz
Phone: +64 9 373 7599 extn 87009

Michael O-Sullivan
Email:  micheael.osullivan@auckland.ac.nz
Phone: +64 9 373 7599 extn 87907

Related publications


Related publications: 

O'Sullivan MJ, Walker CG, Lee D., 2012. Designing Data Storage Tier using Integer Programing, 27th Symposium On Applied Computing, Trento, Italy, 26 Mar 2012 - 30 Mar 2012. Proceedings of 27th Symposium On Applied Computing. ACM. 8 pages.

Lee, D., O'Sullivan, M. J., Walker, C. G., & MacKenzie, M. L. , 2011. Robust Benchmarking for Archival Storage Tiers. In Proceedings of the Sixth workshop on Parallel Data Storage (pp. 6 pages). Seattle, USA.

Lee, D., O'Sullivan, M., & Walker, C., 2011. Measurement for improving the design of commodity archival storage tiers. Proceedings - 2011 4th IEEE International Conference on Utility and Cloud Computing, 275-280.

Lee, D., O'Sullivan, M., & Walker, C., 2011. Benchmarking and modeling disk-based storage tiers for practical storage design. In PMBS'11 - Proceedings of the 2nd International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computing Systems, Co-located with SC'11 (pp. 21-22). Seattle, WA, USA: ACM. doi:10.1145/2088457.2088472

Lee D., O’Sullivan M. J., Walker C. G., 2010. Practical Measurement of Typical Disk Performance and Power Consumption using Open Source SPC-1, Annual International Conference on Green Information Technology – GreenIT, Singapore, Oct 25-26. 

Walker, C.G., O'Sullivan, M.J., 2010. Core-Edge Design of Storage Area Networks - a Single-edge formulation with problem-specific cuts, Computers and Operations Research, 37(5), pp 916-926.

Walker, C.G., O'Sullivan, M.J., Thompson, T.D., 2009. 'A Mixed-Integer Approach to Core-Edge Design of Storage Area Networks', Computers and Operations Research, 34, (10), p2976-3000

O'Sullivan, M. J., & Walker, C. G., 2009. Bi-Criterion Design of Core-Edge Storage Area Networks. In 2009 International Conference on Telecommunications Systems - Modeling and Analysis (pp. 47-71). Monterey.
O'Sullivan, M. J., Shahoumian, T., & Ward, J. M. (2009). US 7502839, Module-building method for designing interconnect fabrics. United States. Retrieved from http://www.patentlens.net/patentlens/patent/US_7502839/en/

Brownlee, J. N., Halytskyy, Y., Jones, N., Kharuk, A. F., O'Sullivan, M. J., Walker, C. G., . . . Ziedins, I., 2009. Storage Network Planning for KAREN/BeSTGRID.

O’Sullivan M.J., Walker C.G., O’Sullivan M.L., Thompson T.D. and Philpott A.B., 2006. Protecting local access telecommunications networks: Toward a minimum-cost solution. Telecommunication Systems Volume 33, pp 353-376.

O’Sullivan M.J., Walker C.G., 2005. A Mixed-integer Approach to Storage Area Network Design using Generic Network Components, School of Engineering Technical Report No. 626. Department of Engineering Science, School of Engineering, University of Auckland, pp1-27.

Walker C.G., O’Sullivan M.J., Elangasinghe M, 2005. Evaluation of Core-Edge Storage Area Network Designs using Simulation, School of Engineering Technical Report No. 627. Department of Engineering Science, School of Engineering, University of Auckland, pp1-24.

 

Top