Page 146 - ITU Kaleidoscope 2016

2016 ITU Kaleidoscope Academic Conference

7. DISCUSSION erything Everywhere (CEE), and Betweenness-Centrality. In
simulations, the CPCE strategy outperformed both CEE and
Throughout this study, the performance of CPCE is assessed Betweenness-Centrality in terms of cache hit rate and stretch
with respect to different parameters, such as popularity model ratio.
variation, cache size, metrics measurement, and topological
impact, i.e., cache deployment. The CPCE caches popular
REFERENCES
contents, and according to Cisco visula networing index [51]
multimedia applications are popular contents because video
on demand (VoD) only from YouTube and Netflex attract [1] G. M. de Brito, P. B. Velloso, and I. M. Moraes, Informa-
tion Centric Networks: A New Paradigm for the Internet.
more than half of the overall traffic, therefore, when the
John Wiley & Sons, 2013.
skewness of the distribution (which is measured through α)
increases, popular contents attain maximum portion of the
[2] V. Jacobson, D. K. Smetters, J. D. Thornton, M. F. Plass,
Internet traffic. N. H. Briggs, and R. L. Braynard, “Networking named
In addition, the current available DRAM is 10GB (maximum), content,” in Proceedings of the 5th international con-
therefore, after some times the memory becomes full and ference on Emerging networking experiments and tech-
content eviction starts for the accommodation of new arrived nologies. ACM, 2009, pp. 1–12.
contents. Consequently, the cache hit rate is reduced and
subsequent requests for the evicted contents are forwarded [3] I. M. Moraes, M. E. M. Campista, L. H. M. Costa,
to the server, which increases the content retrieval delay. To O. C. M. Duarte, J. L. Duarte, D. G. Passos, C. V. N.
add more, with the eviction of contents, the hop decrement is de Albuquerque, and M. G. Rubinstein, “On the impact
affected and thus stretch ratio increases. Similarity, CEE and of user mobility on peer-to-peer video streaming,” IEEE
Betweenness-Centrality also cache diverse kinds of contents Wireless Communications, vol. 15, no. 6, pp. 54–62,
(as they are not particularly designed for popular contents) 2008.
and the popular ones (i.e., VoD) will not have enough space
to be cached. Moreover, in case of memory overflow, there [4] T. Plagemann, V. Goebel, A. Mauthe, L. Mathy,
T. Turletti, and G. Urvoy-Keller, “From content distribu-
is no eviction policy in CEE and Betweenness-Centrality but
tion networks to content networks-issues and challenges,”
they simply delete the contents which affect the mentioned
metrics. Computer Communications, vol. 29, no. 5, pp. 551–562,
2006.
The Cache size is another factor of the CPCE result supremacy,
i.e., when the network becomes stable then CEE and [5] T. Koponen, M. Chawla, B.-G. Chun, A. Ermolinskiy,
Betweenness-Centrality cache sizes overflow and thus evic- K. H. Kim, S. Shenker, and I. Stoica, “A data-oriented
tion operation starts. On the other hand, unlike CEE, the (and beyond) network architecture,” ACM SIGCOMM
CPCE does not cache every content and hence still has space Computer Communication Review, vol. 37, no. 4, pp.
for accommodating new contents rather than deleting the 181–192, 2007.
cached ones. Besides, Betweenness-Centrality stores lesser
contents than the CEE but because of having no eviction [6] J. Kurose, “Content-centric networking: technical per-
policy for Betweenness-Centrality, the popular contents spective,” Communications of the ACM, vol. 55, no. 1,
are evicted and subsequent requests for those contents are pp. 116–116, 2012.
forwarded to the original server. This leads to reduce the
[7] M. Berman, J. S. Chase, L. Landweber, A. Nakao,
performance of Betweenness-Centrality. While in the CPCE
M. Ott, D. Raychaudhuri, R. Ricci, and I. Seskar, “Geni:
the contents are not deleted but cached at other routers, as
A federated testbed for innovative network experiments,”
discussed earlier.
Computer Networks, vol. 61, pp. 5–23, 2014.
8. CONCLUSION [8] FIA, “Future internet assembly (fia),” Tech. Rep.
[Online]. Available: http://www.future-internet.eu/
In the modified draft of ITU - T recommendation [52] it is home/future-internet-assembly.html
requested that every network segment in data aware network-
[9] AsiaFI, “Asiafi,” Tech. Rep. [Online]. Available:
ing (DAN) is recommended to support a caching component
http://www.asiafi.net/meeting/2010/summerschool/
and be additionally ready to assess subscriber requests that
pass through it with the goal that it can make a decision on [10] G. Garcia, A. Beben, F. J. Ramon, A. Maeso, I. Psaras,
subscriber requests and respond using the cached data objects. G. Pavlou, N. Wang, J. Sliwinski, S. Spirou, S. Sour-
To address the ITU-T recommendations, this paper proposes sos et al., “Comet: Content mediator architecture for
a flexible caching strategy, named CPCE, for popular content content-aware networks,” in Future Network & Mobile
caching to improve the performance of ICN caching in terms Summit (FutureNetw), 2011. IEEE, 2011, pp. 1–8.
of cache hit rate and stretch ratio during content download-
ing. The performance of CPCE is compared with two other [11] FP7 Scalable and Adaptive Internet Solutions: SAIL
strategies, i.e., the default ICN strategy, known as Cache Ev- project. [Online]. Available: http://www.sail-project.eu/

– 128 –