Storage device performance prediction with CART models

Authors:
Mengzhi Wang;Kinman Au;Anastassia Ailamaki;Anthony Brockwell;Christos Faloutsos;Gregory R. Ganger
Affiliations:
Carnegie Mellon University;Carnegie Mellon University;Carnegie Mellon University;Carnegie Mellon University;Carnegie Mellon University;Carnegie Mellon University
Venue:
Proceedings of the joint international conference on Measurement and modeling of computer systems
Year:
2004

Citing 0
Cited 10

Storage performance virtualization via throughput and latency control

ACM Transactions on Storage (TOS)
CHAMELEON: a self-evolving, fully-adaptive resource arbitrator for storage systems

ATEC '05 Proceedings of the annual conference on USENIX Annual Technical Conference
Zodiac: efficient impact analysis for storage area networks

FAST'05 Proceedings of the 4th conference on USENIX Conference on File and Storage Technologies - Volume 4
Server-storage virtualization: integration and load balancing in data centers

Proceedings of the 2008 ACM/IEEE conference on Supercomputing
Online monitoring and visualisation of database structural deterioration

International Journal of Autonomic Computing
Workload-aware storage layout for database systems

Proceedings of the 2010 ACM SIGMOD International Conference on Management of data
Improve throughput of storage cluster interconnected with a TCP/IP network using intelligent server grouping

NPC'10 Proceedings of the 2010 IFIP international conference on Network and parallel computing
QBox: guaranteeing I/O performance on black box storage systems

Proceedings of the 21st international symposium on High-Performance Parallel and Distributed Computing
Dynamic global resource allocation in shared data centers and clouds

CASCON '12 Proceedings of the 2012 Conference of the Center for Advanced Studies on Collaborative Research
Model building for dynamic multi-tenant provider environments

ACM SIGOPS Operating Systems Review

Quantified Score

Hi-index	0.00

Visualization

Abstract

This work explores the application of a machine learning tool, CART modeling, to storage devices. We have developed approaches to predict a device's performance as a function of input workloads, requiring no knowledge of the device internals. Two uses of CART models are considered: one that predicts per-request response times (and then derives aggregate values) and one that predicts aggregate values directly from workload characteristics. After training on the device in question, both provide reasonably-accurate black box models across a range of test traces from real environments. An expanded version of this paper is available as a technical report [1].