Self-tuning of disk input-output in operating systems

Authors:
A. Santos;J. Romero;J. Taibo;C. Rodriguez;A. Carballal
Affiliations:
Artificial Neural Networks and Adaptive Systems LAB, University of A Coruña, A Coruña, Spain;Artificial Neural Networks and Adaptive Systems LAB, University of A Coruña, A Coruña, Spain;VideaLAB, University of A Coruña, A Coruña, Spain;Computing and Communications Service, University of A Coruña, A Coruña, Spain;Artificial Neural Networks and Adaptive Systems LAB, University of A Coruña, A Coruña, Spain
Venue:
Journal of Systems and Software
Year:
2012

Citing 7
Cited 0

Automatic TCP buffer tuning

Proceedings of the ACM SIGCOMM '98 conference on Applications, technologies, architectures, and protocols for computer communication
Anticipatory scheduling: a disk scheduling framework to overcome deceptive idleness in synchronous I/O

SOSP '01 Proceedings of the eighteenth ACM symposium on Operating systems principles
A TCP tuning daemon

Proceedings of the 2002 ACM/IEEE conference on Supercomputing
Kernel korner: I/O schedulers

Linux Journal
Self-Learning Disk Scheduling

IEEE Transactions on Knowledge and Data Engineering
Disk schedulers for solid state drivers

EMSOFT '09 Proceedings of the seventh ACM international conference on Embedded software
On disk I/O scheduling in virtual machines

WIOV'10 Proceedings of the 2nd conference on I/O virtualization

Quantified Score

Hi-index	0.00

Visualization

Abstract

One of the most difficult and hard to learn tasks in computer system management is tuning the kernel parameters in order to get the maximum performance. Traditionally, this tuning has been set using either fixed configurations or the subjective administrator's criteria. The main bottleneck among the subsystems managed by the operating systems is disk input/output (I/O). An evolutionary module has been developed to perform the tuning of this subsystem automatically, using an adaptive and dynamic approach. Any computer change, both at the hardware level, and due to the nature of the workload itself, will make our module adapt automatically and in a transparent way. Thus, system administrators are released from this kind of task and able to achieve some optimal performances adapted to the framework of each of their systems. The experiment made shows a productivity increase in 88.2% of cases and an average improvement of 29.63% with regard to the default configuration of the Linux operating system. A decrease of the average latency was achieved in 77.5% of cases and the mean decrease in the request processing time of I/O was 12.79%.