ABSTRACT
Virtualized datacenters strive to reduce costs through workload consolidation. Workloads exhibit a diverse set of IO behaviors and varying IO load that makes it difficult to estimate the IO performance on shared storage. As a result, system administrators often resort to gross overprovisioning or static partitioning of storage to meet application demands. In this paper, we introduce Pesto, a unified storage performance management system for heterogeneous virtualized datacenters. Pesto is the first system that completely automates storage performance management for virtualized datacenters, providing IO load balancing with cost-benefit analysis, per-device congestion management, and initial placement of new workloads.
At its core, Pesto constructs and adapts approximate black-box performance models of storage devices automatically, leveraging our analysis linking device throughput and latency to outstanding IOs.Experimental results for a wide range of devices and configurations validate the accuracy of these models. We implemented Pesto in a commercial product and tested its performance on tens of devices, running hundreds of test cases over the past year. End-to-end experiments demonstrate that Pesto is efficient, adapts to changes quickly and can improve workload performance by up to 19%, achieving our objective of lowering storage management costs through automation.
- Filebench. http://solarisinternals.com/si/tools/filebench/index.php.Google Scholar
- Iometer. http://www.iometer.org.Google Scholar
- G. A. Alvarez, E. Borowsky, S. Go, T. H. Romer, R. Becker-Szendy, R. Golding, A. Merchant, M. Spasojevic, A. Veitch, and J. Wilkes. Minerva: An Automated Resource Provisioning Tool for Large-Scale Storage Systems. In ACM Transactions on Computer Systems, Nov. 2001. Google ScholarDigital Library
- E. Anderson. Simple table-based modeling of storage devices. Technical report, HPL-SSP-2001-4, HP Labs, July 2001.Google Scholar
- E. Anderson, M. Hobbs, K. Keeton, S. Spence, M. Uysal, and A. Veitch. Hippodrome: running circles around storage administration. In Proceedings of the 1st USENIX conference on File and Storage Technologies, FAST'02, Berkeley, CA, USA, 2002. USENIX Association. Google ScholarDigital Library
- S. Chen and D. Towsley. The Design and Evaluation of RAID 5 and Parity Striping Disk Array Architectures. Journal on Parallel and Distributed Computing, 17(1--2):58--74, 1993. Google ScholarDigital Library
- S. Chen and D. Towsley. A performance evaluation of RAID architectures. IEEE Transactions on Computers, 45:1116--1130, 1996. Google ScholarDigital Library
- T. E. Denehy, J. Bent, F. I. Popovici, A. C. Arpaci-Dusseau, and R. H. Arpaci-Dusseau. Deconstructing storage arrays. In Proceedings of the 11th international conference on Architectural support for programming languages and operating systems, ASPLOS-XI, pages 59--71, New York, NY, USA, 2004. ACM. Google ScholarDigital Library
- G. Ganger. Automated disk drive characterization. http://www.pdl.cmu.edu/Dixtrac/index.shtml.Google Scholar
- C. C. Gotlieb and G. H. MacEwen. Performance of Movable-Head Disk Storage Devices. Journal of the ACM, 20(4):604--623, 1973. Google ScholarDigital Library
- A. Gulati, I. Ahmad, and C. A. Waldspurger. PARDA: Proportional allocation of resources for distributed storage access. In Proceedings of the 7th conference on File and Storage Technologies, pages 85--98, Berkeley, CA, USA, 2009. USENIX Association. Google ScholarDigital Library
- A. Gulati, C. Kumar, and I. Ahmad. Storage Workload Characterization and Consolidation in Virtualized Environments. In Workshop on Virtualization Performance: Analysis, Characterization, and Tools (VPACT), 2009.Google Scholar
- A. Gulati, C. Kumar, I. Ahmad, and K. Kumar. BASIL: Automated IO load balancing across storage devices. In Proceedings of the 8th USENIX conference on File and Storage Technologies, FAST'10, Berkeley, CA, USA, 2010. USENIX Association. Google ScholarDigital Library
- J. L. Hennessy and D. A. Patterson. Computer Architecture: A Quantitative Approach, Fourth edition. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 2007. Google ScholarDigital Library
- R. Jain and I. Chlamtac. The P2 algorithm for dynamic calculation of quantiles and histograms without storing observations. Communications of the ACM, 28:1076--1085, October 1985. Google ScholarDigital Library
- T. Kelly, I. Cohen, M. Goldszmidt, and K. Keeton. Inducing models of black-box storage arrays. Technical Report HPL-2004-108, HP Labs, 2004.Google Scholar
- M. Kim and A. Tantawi. Asynchronous disk interleaving: Approximating access delays. IEEE Transactions on Computers, 40(7):801--810, 1991. Google ScholarDigital Library
- E. K. Lee and R. H. Katz. An analytic performance model of disk arrays. SIGMETRICS Performance Evaluation Review, 21(1):98--109, 1993. Google ScholarDigital Library
- J. D. C. Little. A Proof for the Queuing Formula: L = λW. Operations Research, 9(3), 1961.Google Scholar
- A. Mashtizadeh, E. Celebi, T. Garfinkel, and M. Cai. The Design and Evolution of Live Storage Migration in VMware ESX. In Proc. USENIX Annual Technical Conference (ATC '11), June 2011. Google ScholarDigital Library
- A. Merchant and P. S. Yu. An analytical model of reconstruction time in mirrored disks. Performance Evaluation, 20:115--129, May 1994. Google ScholarDigital Library
- A. Merchant and P. S. Yu. Analytic Modeling of Clustered RAID with Mapping Based on Nearly Random Permutation. IEEE Transactions on Computers, 45(3), 1996. Google ScholarDigital Library
- D. R. Merrill. Storage economics: Four principles for reducing total cost of ownership. May 2009. http://www.hds.com/assets/pdf/four-principles-for-reducing-total-cost-of-ownership.pdf.Google Scholar
- M. P. Mesnier, M. Wachs, R. R. Sambasivan, A. X. Zheng, and G. R. Ganger. Modeling the relative fitness of storage. SIGMETRICS Performance Evaluation Review, 35(1), 2007. Google ScholarDigital Library
- J. D. Padhye, A. L. Rahatekar., and L. W. Dowdy. A Simple LAN File Placement Strategy. In International CMG Conference, 1995.Google Scholar
- C. Ruemmler and J. Wilkes. An introduction to disk drive modeling. IEEE Computer, 27:17--28, 1994. Google ScholarDigital Library
- E. Shriver, A. Merchant, and J. Wilkes. An Analytic Behavior Model for Disk Drives with Readahead Caches and Request Reordering. SIGMETRICS Performance Evaluation Review, 26(1):182--191, 1998. Google ScholarDigital Library
- N. Simpson. Building a data center cost model. Jan 2010. http://www.burtongroup.com/Research/DocumentList.aspx?cid=49.Google Scholar
- E. Thereska, M. Abd-El-Malek, J. J. Wylie, D. Narayanan, and G. R. Ganger. Informed data distribution selection in a self-predicting storage system. In International Conference on Autonomic Computing, 2006. Google ScholarDigital Library
- E. Thereska and G. R. Ganger. IRONModel: Robust performance models in the wild. SIGMETRICS Performance Evaluation Review, 36:253--264, June 2008. Google ScholarDigital Library
- A. Thomasian and J. Menon. Performance analysis of RAID5 disk arrays with a vacationing server model for rebuild mode operation. In Proceedings of the Tenth International Conference on Data Engineering, pages 111--119. IEEE Computer Society, 1994. Google ScholarDigital Library
- M. Uysal, G. A. Alvarez, and A. Merchant. A modular, analytical throughput model for modern disk arrays. In IEEE International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunications Systems (MASCOTS), 2001. Google ScholarDigital Library
- E. Varki, A. Merchant, J. Xu, and X. Qiu. Issues and challenges in the performance analysis of real disk arrays. IEEE Transactions on Parallel and Distributed Systems, 15:559--574, 2004. Google ScholarDigital Library
- VMware, Inc. VMware Storage VMotion: Non-Disruptive, Live Migration of Virtual Machine Storage, 2007. http://vmware.com/files/pdf/storage_vmotion_datasheet.pdf.Google Scholar
- VMware, Inc. vSphere Resource Management Guide: ESX 4.1, ESXi 4.1, vCenter Server 4.1. 2010.Google Scholar
- VMware, Inc. VMware vSphere. 2011. http://www.vmware.com/products/vsphere/overview.html.Google Scholar
- VMware, Inc. VMware vStorage VMFS. 2011. http://www.vmware.com/files/pdf/VMware-vStorage-VMFS-DS-EN.pdf.Google Scholar
- T. Voellm. Useful IO profiles for simulating various workloads. http://blogs.msdn.com/b/tvoellm/archive/2009/05/07/useful-io-profiles-for-simulating-various-workloads.aspx.Google Scholar
- M. Wang, K. Au, A. Ailamaki, A. Brockwell, C. Faloutsos, and G. R. Ganger. Storage Device Performance Prediction with CART Models. In IEEE International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunications Systems (MASCOTS), pages 588--595, 2004. Google ScholarDigital Library
- P. S. Yu and A. Merchant. Analytic modeling and comparisons of striping strategies for replicated disk arrays. IEEE Transactions on Computers, 44:419--433, March 1995. Google ScholarDigital Library
Index Terms
- Pesto: online storage performance management in virtualized datacenters
Recommendations
Romano: autonomous storage management using performance prediction in multi-tenant datacenters
SoCC '12: Proceedings of the Third ACM Symposium on Cloud ComputingWorkload consolidation is a key technique in reducing costs in virtualized datacenters. When considering storage consolidation, a key problem is the unpredictable performance behavior of consolidated workloads on a given storage system. In practice, ...
Enabling Instantaneous Relocation of Virtual Machines with a Lightweight VMM Extension
CCGRID '10: Proceedings of the 2010 10th IEEE/ACM International Conference on Cluster, Cloud and Grid ComputingWe are developing an efficient resource management system with aggressive virtual machine (VM) relocation among physical nodes in a data center. Existing live migration technology, however, requires a long time to change the execution host of a VM, it ...
Virtual Machine Migration Method between Different Hypervisor Implementations and Its Evaluation
WAINA '12: Proceedings of the 2012 26th International Conference on Advanced Information Networking and Applications WorkshopsVirtualization technologies are an important building block for cloud services. Each service will run on virtual machines (VMs) deployed over different hyper visors in the future. Therefore, a VM migration method between different hyper visor ...
Comments