You are on page 1of 2

2011 19th IEEE International Conference on Network Protocols

IPv6 Evolution, Stability and Deployment


Xiaoke Jiang , Jun Bi† , Yangyang Wang , Zhijie He , Wei Zhang , Hongchen Tian
{ Dept. of Computer Science, † Network Research Center}, Tsinghua University, Beijing, 100084, P.R.C


Email: {justok, wyy, hzj, zw, tianhc}@netarchlab.tsinghua.edu.cn, † junbi@tsinghua.edu.cn

Abstract—Our subject focuses on IPv6 network, which devel- 8000 35 0.17


AS# Prefix
ops for more than 10 years. How IPv6 evolve in those years? Is Link#
Prefix#
PATH axes x1y1
Coef.(y2 axis)
7000
IPv6 network mature enough to undertake the load produced by 30 0.16

users? Can we find some principles to guide IPv6 deployment, 6000

25 0.15
which make the whole network more robust and efficiency? 5000

This paper tries to answer these questions with in-depth 4000 20 0.14

statistics. Good news is that network is growing at a speed of


O(d2 ) (d is time) after 2006, moreover, network itself and its
3000
15 0.13

routing system become more and more stable. And we explore 2000

10 0.12
special properties of this preliminary network, We find that 1000

distribution of AS degree follows "Power-Law Distribution", but 0 5 0.11


2003 2004 2005 2006 2007 2008 2009 2010 2011 2003 2004 2005 2006 2007 2008 2009 2010 2011
AS-level topology cannot be described as "Small-World Model"
properly. We also propose a method to define the importance (a) Amount of Entities (b) Avg Life, Cluster Coef
of AS and give a simple principle of IPv6 deployment. We even Fig. 1: Evolution of IPv6 Network
build "6Stats Project"[1] to provide data which help deploy IPv6.

Index Terms—IPv6, Routing, Deployment, Network Measure- get similar result. So O(d2 ) is better than O(ed ) to describe
ment
network size. This matches people’s intuition: changing from
I. I NTRODUCTION O(d) to O(d2 ) is smoother than to O(ed ).
IPv6 experienced a slow development for a long time since B. Special Properties of Preliminary Network
1999, but now it goes into a rapid growth stage, which is
According to our statistics, amount of IPv6 prefix and ASes
different from IPv4[4].
is just 1% of those of IPv4, we suppose IPv6 is not ready to
But most of research on IPv6 uses a poor dataset, which may
undertake all the traffic people produce daily.
be small or outdated. What’s more, most of existing research
And we also measure the average AS_PATH length from
follows what we have done on IPv4. As a newborn technology,
monitors to other AS. We find an abnormal phenomenon.
the special properties haven’t get enough attention. So we
Small-World Model infer that length increase slowly while
calculate a lot of indexes to show its tomography, based on
amount of node increase, which is positive correlation. But
which we study on principles of IPv6 deployment using graph
from 2003 to 2010, amount of AS increase(Fig 1a) while
theory.
average length decrease! Fig 3a shows this trend 1 . That is
In this paper, we use whole IPv6 BGP RIB data collect by
negative correlation. So Small-World Model cannot describe
RouteViews Project from May 3/2003 to Oct 31/2010, Taking
IPv6 AS-level topology. But Log-log CCDF line in Fig 3b is
account of limitation of monitors and this methodology itself,
nearly liner, which proves that AS degree meets Power-Law
we mostly calculate one index result instance using data of
Distribution.
whole month.
We suppose the phenomenon is result of small ASes and
II. E VOLUTION links set, which leading to no "shortcuts" for packet forward-
A. Growing at a Speed of O(d2 ) ing between AS pairs. So packet has to follow a long path
without choice.
Fig 1a gives amount of prefixes, ASes, links. If "d" rep-
resents time gap between statistics date and starting date, III. S TABILITY OF IP V 6 N ETWORK
network grew with the speed of O(d) before 2006[3], And A. Key Indexes
after [4] argues the speed is O(ed ). But we suppose O(d2 ) is
more accurate. Fig 1b shows average prefix life, path life and cluster
We draw asymptotic lines of those results between Jan/2006 coefficient in one month. Life here means appearing days per
and Oct/2010 with quadratic polynomial and exponential. For months. Path life obviously becomes longer. Upper bound of
link line(Highest, Green), prefix life doesn’t change clearly, but lower bound increases.
As for cluster coefficient, it shook more fiercely before 2006
y = 0.0025x2 − 196.36x + 106 (R2 = 0.9927) than after that.
is better than y = 3×10−11 e0.0008x (R2 = 0.9445) , for 0.99 1 here path is AS_PATH from other ASes to RouteViews monitors. So result
is bigger than 0.94. As to AS(Lowest, Red) and prefix line, we here is not average path length of whole network, but it does show the trend.

978-1-4577-1394-1/11/$26.00 ©2011 IEEE 123


3.5
month
week
day similar index, clustering coefficient[2] (short as coef in the
3 following), which reflects node trend of clustering, or "friends
of friends".
2.5
Pick 3 types of ASes first: ASes with most neighbors,
2
biggest local coef and smallest local coef, excluding coef equal
0 or 1 in order to clear effect of single link and small total joint
1.5 group. Then recalculate average coef after removing AS in the
picked list one by one. We find that removal of ASes with
1
2003 2004 2005 2006 2007 2008 2009 2010 2011
most neighbors cause greatest effect on the new average coef.
Fig. 2: Path# for AS Pair in one Month/Week/Day This proves that kind of ASes own more important position
in AS-level topology, for they contribute more to network
5.8 1 10 100 1000
performance.
Average Length of AS PATH 1 1

5.6
CDF
Log-Log CCDF (x2-y2) So we suppose that there does exist strategies which help to
5.4 0.9
increase the robust and efficient of network when deploy IPv6.
5.2
0.8 0.1
For example, connecting to AS "important" neighbors, which
5

4.8
leads average path to other ASes short and that AS forwarding
0.7

4.6
efficiency(but the whole network forwards may not be the most
4.4
0.6 0.01
efficiency). And other deeply strategies can be explored.
4.2
0.5

4 V. C ONCLUSIONS
3.8
2003 2004 2005 2006 2007 2008 2009 2010 2011
0.4
0 100 200 300 400 500 600 700 800 900
0.001
In this paper, our conclusions include:
(a) average path length (b) AS Degree CDF and log-log CCDF • Scale of IPv6, growing at a speed of O(d2 ), not O(ed )
Fig. 3: Special Phenomenons as some paper argues. Growing doesn’t follow Moore’s Law,
but does go into rapid spread stage.
We also analyze appearance and disappearance of path, links • IPv6 network itself and its Routing System are becoming
and AS, prefix length and origin AS of prefix, all indexes show more and more stable.
trend of becoming stable. In particular, status after 2006 is • Year 2006 is a milestone of network. After 2006, network
more stable than before that, Fig 3a is an evidence. is more mature and stable and develops faster. We suppose
B. Instability of Routing System the reason is that network change gradually from a scientific
research network to production network, for example, applica-
There are 3 lines in Fig 2, which represent average amount
tions of CNGI-CERNET2 in China, IPv6 site of Google and
of paths for AS pairs(Source AS and destination AS) calcu-
IPv6 commercial service of Comcast.
lated in different intervals: Month, Week and Day. The value
• AS-level topology cannot described as Small-World
decreases over time, especially the month line, decreasing from
method, which reflects properties of its preliminary stage.
3 in 2003 to 1.5 in 2010.
• AS with more neighbors take more importance position
We also explore the similarities of set of AS, path and links
in the topology. Connecting with importance AS is a good
between sequential months, and the similarities increase over
strategy of establishing new link, which should get special care
time. All those prove that network itself and its routing system
when nearly half of ASes access the Internet by single Link.
are becoming stable. And year 2006 is prove to be milestone
And our future work will explore more special properties, a
of stability, too.
more accurate model to define importance of AS and give
IV. D ISCUSSION OF D EPLOYMENT more deeply principles to guide IPv6 deployment.
A. Environment Faced VI. ACKNOWLEDGMENTS
Fig 3b is AS degree CDF and log-log CCDF graph2 , which Supported by National Science Foundation of China un-
shows the environment ISP facing to deploy IPv6. 44% of der Grant 61073172. Program for New Century Excellent
ASes only have one degree. CDF grows very fast at start-up, Talents in University. Specialized Research Fund for the
more than 80% of ASes have degree small than 6. It grows Doctoral Program of Higher Education of China under Grant
very slow at later stage, less than 8% of ASes have degree 200800030034,
more than 16, but maximal degree is more than 800.
R EFERENCES
B. Importance of AS [1] 6Stats Project. http://202.112.49.243/Web6/index/.
We suppose that the best way to measure importance of [2] Cluster Coefficient. http://en.wikipedia.org/wiki/Clustering_coefficient.
[3] W. Eddy. Basic properties of the ipv6 as-level topology. ACM SIGMET-
AS is its impact to the average shortest path length, but this RICS Performance Evaluation Review, 36(3):50–57, 2008.
is really hard under the circumstance of complicated BGP [4] G. Zhang, B. Quoitin, and S. Zhou. Phase changes in the evolution of
configuration and business relationship. So we pick another the ipv4 and ipv6 as-level internet topologies. Computer Comunications
(COMCOM), 34(5):649–657, 2011.
2 Based on data in Mar/2010, others show similar properties

124

You might also like