Next: Contents
Contents
The Effect of Deceptive Idleness
on Disk Schedulers
Sitaram Iyer
Abstract
Disk schedulers in operating systems are
generally work-conserving; they schedule a request immediately after the
previous request has finished. Such schedulers need multiple outstanding
requests to make good decisions. Unfortunately, many applications issue
synchronous, almost-continuous streams of read requests. This forces the
scheduler into making decisions too early, falsely assuming that the process
has become momentarily idle. This phenomenon of deceptive idleness
causes significant degradation in performance and quality of service
objectives on current systems. We solve deceptive idleness by designing and
implementing a transparent, non-work-conserving scheduling framework
for various scheduling policies. We evaluate this solution on
microbenchmarks and real workloads, and observe large benefits. The Apache
webserver delivers 56% and 16% more throughput for two configurations. The
Andrew Benchmark runs faster by 8% (54% for the read-intensive phase).
Variants of the TPC-B database benchmark exhibit improvements between 4%
and 60%. Proportional-share schedulers become empowered to efficiently
deliver application-desired proportions.
Acknowledgements
I wish to thank a number of people who have contributed
to the making of this thesis. First and foremost, I thank my advisor Dr.
Peter Druschel for his visionary supervision throughout the project, and the
freedom and flexibility that he allowed me at every stage. The driving force
can be rightly attributed to his adamant insistence, especially during the
formative days, that this project can actually be pulled off. I cannot
sufficiently thank Juan Navarro and Karthick Rajamani for the marvellous
feedback-based sounding boards that they have both been. Mohit Aron gave me
useful advice regarding the choice and usage of timing mechanisms. I had
countless lunchtime conversations with Sanjeeb Dash about the material in
this thesis. Martin DeMello coined the phrase ``deceptive idleness'' after
hearing a boring description of the same; it sounds so much cuter than the
originally conceived ``apparent idleness''. Sameer Siruguri helped print and
submit the thesis when I wasn't physically around. Both Rice University and
MIT have provided computational resources used for this project; I thank
them for that. Finally, I express sincere thanks to my family for their
continued support over the years.
Sitaram Iyer ssiyer@cs.rice.edu