The Minion architecture of Nowlan et
al. 12 has several features that would be
beneficial and complementary to Paceline’s strengths, such as its SSL strategy
and wire-compatible changes to TCP
that allow unordered receive.
Paceline reduces sender-side queuing delay by writing the minimum
amount to the TCP socket buffer. Recently, the bufferbloat problem has received greatly renewed general attention
and interest. The recent algorithm work
by Kathleen Nichols and Van Jacobson
is a promising step since it is easy to implement and needs no configuration. 11
What is still missing, however, is true
end-to-end evaluation that quantifies
the total combined effect of refinements
to TCP, ECN (Explicit Congestion Notification), AQM (Active Queue Management), and adaptive multimedia.
Underlying our interest is what we
believe remains an open question: Is
there really a line between applications
that needs to break from TCP for interactivity reasons? This is a line that has
been moving steadily over time.
Lessons Learned
Introducing a new transport layer is
challenging because the designer must
address several critical issues. First, the
transport layer has to improve performance significantly for the target applications in order to justify the extra effort. Second, the enhancements should
not negatively affect other traffic types;
we adhere to the vision that the Internet
should remain a general-purpose infrastructure for a vast array of applications.
Finally, the performance improvements
at the transport level in terms of latency,
fairness, and utilization need to translate to quality improvements at the application level. Paceline has shown positive results in all these aspects.
We tested Paceline extensively using videoconferencing scenarios over
WAN settings. Paceline was also used
in a small cloud-based game prototype
to scale the wide area communication
of an Epic-scale game scenario. 13 Both
of these applications show significant
improvements in multimedia quality because of the reductions in TCP
sender-side delays and the use of adaptation in Paceline.
To ensure we have a general-purpose
transport supporting a wide range of
applications, we tested using a network
traffic generator simulating different
HTTP Web traffic flavors (for example,
HTTP1.0 and HTTP1.1, with and with-
out pipelining). The experiment tested
Web document downloads while vary-
ing the number of objects per page and
the page size. Paceline improved the
bandwidth utilization using the multi-
streaming feature as we increased the
number of objects as a result of the
automatic pipelining of small transac-
tions on top of the underlying channel.
acknowledgments
Thanks to colleagues at the Network,
Systems, and Security lab at the Univer-
sity of British Columbia for their feed-
back on previous drafts of the article.
Special thanks to Mahdi Tayarani Na-
jaran for his significant contributions
to the design and implementation of
Paceline. Finally, thanks to Terry Coatta
for helping to improve both the style
and content of this article.
Related articles
on queue.acm.org
Four Billion Little Brothers?: Privacy, mobile
phones, and ubiquitous data collection
Katie Shilton
http://queue.acm.org/detail.cfm?id=1597790
VoIP: What is it Good for?
Sudhir R. Ahuja and Robert En
http://queue.acm.org/detail.cfm?id=1028897
Data in Flight
Julian Hyde
http://queue.acm.org/detail.cfm?id=1667562
References
1. Bharambe, A., douceur, J. R., Lorch, J.R., moscibroda,
T., Pang, J., Seshan, S. and Zhuang, X. donnybrook:
Enabling large-scale, high-speed, peer-to-peer games.
In Proceedings of the ACM SIGCOMM Conference on
Data Communication, (2008), 389–400.
2. The Chromium Projects. SPd Y: An experimental
protocol for a faster Web, (2011); http://www.
chromium.org/spdy/spdy-whitepaper.
3. Erbad, A. Real-time support for interactive multimedia
applications (2012); http://hdl.handle.net/2429/42878.
4. Erbad, A., Hutchinson, N.C. and Krasic, C. dOHA:
Scalable real-time Web applications through adaptive
concurrent execution. In Proceedings of the 21st
International Conference on World Wide Web, (2012),
161–170.
5. Erbad, A., Tayarani Najaran, m. and Krasic, C. Paceline:
latency management through adaptive output.
In Proceedings of the First Annual ACM SIGMM
Conference on Multimedia Systems, (2010), 181–192.
6. Ford, B. Structured streams: a new transport
abstraction. In Proceedings of the Conference on
Applications, Technologies, Architectures, and Protocols
for Computer Communications, (2007), 361–372.
7. gettys, J. and Nichols, K. Bufferbloat: dark buffers in
the Internet. Queue 9, 11 (2011), 40:40–40: 54.
8. goel, A., Krasic, C. and Walpole, J. Low-latency
adaptive streaming over TCP. ACM Transactions
on Multimedia Computing, Communications, and
Applications 4, 3 (2008), 1–20.
9. Krasic, C. A framework for quality-adaptive media
streaming: Encode once–stream anywhere. Ph.d.
thesis. AAI3119036 (2004).
10. Kuschnig, R., Kofler, I. and Hellwagner, H. An
evaluation of TCP-based rate-control algorithms
for adaptive Internet streaming of H.264/SVC.
In Proceedings of the First Annual ACM SIGMM
Conference on Multimedia Systems, (2010), 157–168.
11. Nichols, K. and Jacobson, V. Controlling queue delay.
Queue 10, 5 (2012), 20:20–20: 34.
12. Nowlan, m., Tiwari, N., Iyengar, J., Amin, S. and
Ford, B. Fitting square pegs through round pipes:
unordered delivery wire compatible with TCP and TLS.
In Proceedings of the 9th Conference on Networked
Systems Design and Implementation, (2012).
13. Tayarani Najaran, m. and Krasic, C. Scaling online
games with adaptive interest management in the
cloud. In Proceedings of the 9th Annual Workshop on
Network and Systems Support for Games, (2010),
9:1–9: 6.
Aiman Erbad recently joined the computer science and
engineering department at qatar University as an assistant
professor after completing his Ph.d. at the University
of British Columbia. His research interests include Web
architecture, networking, real-time multimedia, ubiquitous
computing, and concurrency support.
Charles “Buck” Krasic works on large-scale processing
of video at google’s You Tube. Prior to joining You Tube, he
was a professor in the department of computer science
at the University of British Columbia. His research led to
the development of qStream. Other research interests
include multimedia, operating systems, networking, and
distributed systems.