Application of both Temporal and Spatial Localities in the Management of Kernel Buffer Cache - Advanced Operating Systems and Kernel Applications - page 117

including kernel of OS, would be of great potential
for improving I/O performance and removing the
painful I/O bottleneck.A challenge in the approach
is how to make it reconcile with the virtualization
technologies and effectively deal with portability
issues, which all require isolation from the low-
level details to some extent.

Black, D., Carter, J., Feinberg, G., MacDonald,
R., Mangalat, S., Sheinbrood, E., et al. (1991).
OSF/1 Virtual Memory Improvements . USENIX
Mac Symposium.

Cao, P., & Irani, S. (1997). Cost-Aware WWW
Proxy Caching Algorithms . USENIX Annual
Technical Conference.

Ding, X., Jiang, S., Chen, F., Davis, K., & Zhang,
X. (2007). DiskSeen: Exploiting Disk Layout and
Access History to Enhance I/O Prefetch . USENIX
Annual Technical Conference.

concluSion

In this chapter, we identify a serious weakness of
lacking spatial locality exploitation in I/O caching,
and propose a new and effective memory man-
agement scheme, DULO, which can significantly
improve I/O performance by exploiting both tem-
poral and spatial locality. Our experiment results
show that DULO can effectively reorganize ap-
plication I/O request streams mixed with random
and sequential accesses in order to provide a more
disk-friendly request stream with high sequenti-
ality of block accesses. We present an effective
DULO replacement algorithm to carefully tradeoff
random accesses with sequential accesses and
evaluate it using traces representing representative
access patterns. The results of experiments on a
prototype implementation of DULO in a recent
Linux kernel show that DULO can significantly
improve the I/O performance for many applica-
tions from different areas. As DULO represents
a promising effort in removing the I/O barrier for
many I/O-intensive applications, more research in
this direction is called to reveal its full potential
and address its issues.

Ganger, G., & Kaashoek, F. (1997). Embedded
Inodes and Explicit Groups: Exploiting Disk
Bandwidth for Small Files. USENIX Annual
Technical Conference.

Hsu, W. W., Young, H. C., & Smith, A. J. (2003).
The Automatic Improvement of Locality in Stor-
age Systems . Technical Report CSD-03-1264,
UC Berkeley.

Iyer, S., & Druschel, P. (2001). Anticipatory
Scheduling: A Disk Scheduling Framework to
Overcome Deceptive Idleness in Synchronous
I/O. 18th ACM Symposium on Operating Systems
Principles .

Jiang, S., Chen, F., & Zhang, X. (2005). CLOCK-
Pro: An Effective Improvement of the CLOCK
Replacement . USENIX Annual Technical Con-
ference.

Johnson, T., & Shasha, D. (1994). 2Q: A Low
Overhead High Performance Buffer Management
ReplacementAlgorithm. In International Confer-
ence on Very Large Data Bases , (pp. 439-450).

referenceS

Mckusick, M. K., Joy, W. N., Leffler, S. J., &
Fabry, R. S. (1884). A Fast File System for UNIX.
Transactions on Computer Systems, 2 (3).

Arpaci-Dusseau, R. H., Arpaci-Dusseau, N. C.,
Burnett, T. E., Denehy, T. J., Engle, H. S., Gunawi,
J., & Nugent, F. I. Popovici. (2003). Transform-
ing Policies into Mechanisms with Infokernel .
19th ACM Symposium on Operating Systems
Principles.

Schindler, J., Griffin, J. L., Lumb, C. R., & Ganger,
G. R. (2002). Track-Aligned Extents: Matching
Access Patterns to Disk Drive Characteristics .
USENIX Conference on File and Storage Tech-
nologies.