Relative fitness models for storage

ABSTRACT

Relative fitness is a new black-box approach to modeling storage devices. Whereas conventional black-box models train to predict a device's performance given "device-independent" workload characteristics, relative fitness models learn to predict the changes in performance between specific devices. There are two advantages. First, unlike conventional modeling, relative fitness does not depend entirely on workload characteristics; performance and resource utilization (e.g., cache usage) can also be used to describe a workload. This is beneficial when workload characteristics are difficult to express (e.g., temporal locality). Second, because relative fitness models are constructed for each pair of devices, changes in workload characteristics (e.g., I/O inter-arrival delay) can be modeled. Therefore, unlike a conventional model, a relative fitness model can be used by applications with a closed I/O arrival process. In this article, we present relative fitness as an evolution of the conventional model and share some early results.

REFERENCES

• 1: Guillermo A. Alvarez , Elizabeth Borowsky , Susie Go , Theodore H. Romer , Ralph Becker-Szendy , Richard Golding , Arif Merchant , Mirjana Spasojevic , Alistair Veitch , John Wilkes, Minerva: An automated resource provisioning tool for large-scale storage systems, ACM Transactions on Computer Systems (TOCS), v.19 n.4, p.483-518, November 2001 [doi>10.1145/502912.502915]
• 2: E. Anderson. Simple table-based modeling of storage devices. SSP Technical Report HPL-SSP-2001-4. HP Laboratories, July 2001.
• 3: Eric Anderson , Michael Hobbs , Kimberly Keeton , Susan Spence , Mustafa Uysal , Alistair C. Veitch, Hippodrome: Running Circles Around Storage Administration, Proceedings of the Conference on File and Storage Technologies, p.175-188, January 28-30, 2002
• 4: C. Borgelt. A Decision Tree Plug-In for DataEngine. www.netapp.com.
• 5: E. Borowsky, et al. Using attribute-managed storage to achieve QoS. International Workshop on Quality of Service. IFIP, 1997.
• 6: G. R. Ganger. Generating representative synthetic workloads: an unsolved problem. International Conference on Management and Performance Evaluation of Computer Systems, 1995.
• 7: Gregory R. Ganger , Yale N. Patt, Using System-Level Models to Evaluate I/O Subsystem Designs, IEEE Transactions on Computers, v.47 n.6, p.667-678, June 1998 [doi>10.1109/12.689646 ]
• 8: Intel. iSCSI. www.sourceforge.net/projects/intel-iscsi.
• 9: Zachary Kurmas, Using the Distiller to Direct the Development of Self-Configuration Software, Proceedings of the First International Conference on Autonomic Computing (ICAC'04), p.172-179, May 17-18, 2004
• 10: Z. Kurmas, et al. Synthesizing Representative I/O Workloads Using Iterative Distillation. International Workshop on Modeling, Analysis, and Simulation of Computer and Telecommunications Systems. IEEE/ACM, 2003.
• 11: J. Satran. iSCSI. www.ietf.org/rfc/rfc3720.txt.
• 12: Transaction Processing Performance Council. TPC Benchmark C, December 2002. www.tpc.org/tpcc/ Revision 5.1.0.
• 13: Mengzhi Wang , Kinman Au , Anastassia Ailamaki , Anthony Brockwell , Christos Faloutsos , Gregory R. Ganger, Storage Device Performance Prediction with CART Models, Proceedings of the The IEEE Computer Society's 12th Annual International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunications Systems (MASCOTS'04), p.588-595, October 04-08, 2004

INDEX TERMS

Primary Classification:
  B. Hardware
  B.3 MEMORY STRUCTURES
      B.3.2 Design Styles
          Subjects: Mass storage (e.g., magnetic, optical, RAID)

Additional Classification:
  C. Computer Systems Organization
  C.4 PERFORMANCE OF SYSTEMS
      Subjects: Reliability, availability, and serviceability; Modeling techniques

  E. Data
  E.5 FILES

  H. Information Systems
  H.3 INFORMATION STORAGE AND RETRIEVAL
      H.3.2 Information Storage

General Terms:
Design, Management