Inter-Domain QoS Routing Scheme using Link State Information

Abstract

In the Internet, in order to guarantee Quality of Service (QoS), all routers have to manage their Link State Databases (LSDBs) and determine a routing path that satisfies service requirements based on the LSDB. In addition, routing protocols have to provide QoS not only in the intra- domain but also in the inter-domain to enable the end to end QoS guarantee. In the intra-domain routing, it is possible that a router manages its LSDB because the link state information is small in a domain. On the other hand, in inter- domain routing, since the amount of the link state information is very large, it is unrealistic to manage the whole link state information. The Border Gateway Protocol (BGP) is the most prominent inter-domain routing protocol in the current internet. However, it is difficult to guarantee the inter-domain QoS using BGP because the BGP is a path- vector based protocol. In this paper, we propose the inter- domain QoS routing schemes that can manage the LSDBs effectively and minimize a waste of the network system resources by Link State Update (LSU) massages. We evaluate the performance of the proposed schemes by simulations. From the results, the proposed schemes can reduce the blocking probability and amount of the link state update with a few link state information.

Full text

Page 1

Abstract - In the Internet, in order to guarantee Quality
of Service (QoS), all routers have to manage their Link State
Databases (LSDBs) and determine a routing path that
satisfies service requirements based on the LSDB. In addition,
routing protocols have to provide QoS not only in the intra-
domain but also in the inter-domain to enable the end to end
QoS guarantee. In the intra-domain routing, it is possible
that a router manages its LSDB because the link state
information is small in a domain. On the other hand, in inter-
domain routing, since the amount of the link state
information is very large, it is unrealistic to manage the
whole link state information. The Border Gateway Protocol
(BGP) is the most prominent inter-domain routing protocol
in the current internet. However, it is difficult to guarantee
the inter-domain QoS using BGP because the BGP is a path-
vector based protocol. In this paper, we propose the inter-
domain QoS routing schemes that can manage the LSDBs
effectively and minimize a waste of the network system
resources by Link State Update (LSU) massages. We evaluate
the performance of the proposed schemes by simulations.
From the results, the proposed schemes can reduce the
blocking probability and amount of the link state update with
a few link state information.
Keywords - QoS routing; link state data base; blocking
probability; update rate.

I. Introduction
In recent years, as the internet user population expands
rapidly, the demands of real-time multimedia service over the
internet increase. However, since the traditional internet
supports best effort services, it is difficult to satisfy the
demands of real-time service. In order to guarantee this
requirement, the QoS metric such as bandwidth, delay, jitter,
packet loss, etc., should be controlled properly[1][2].
The internet consists of many autonomous systems
(ASs). Routing schemes are divided into routing inside an AS
and between ASs which are referred to as intra-domain
routing and inter-domain routing, respectively[3]. Although
much research for QoS guarantee has been conducted, the
majority of the works were limited in intra-domain routing
and little work for inter-domain routing has been done[4]. To
enable the end to end QoS, routing protocols have to provide
QoS in the both intra-domain and inter-domain. The BGP is
the de facto standard for the inter-domain routing. However,
because the BGP is a path-vector based protocol, it is difficult
to guarantee the inter-domain QoS using the BGP[5].
In order to guarantee QoS, all routers have to manage
their Link State Databases (LSDBs) and determine the
routing path that satisfies service requirements based on the
LSDBs. In the intra-domain routing, it is possible that a
router manages its LSDB because the link state information
is small in only one AS. In the inter-domain routing, because
the routing range is much larger than the intra-domain routing,
amount of the managed link state information dramatically
increases. In addition, network system resources are
excessively wasted by transmitting Link State Update (LSU)
messages[6]. Therefore, it is unrealistic for a router to
manage the link state information of all ASs for the inter-
domain QoS routing.
We propose the inter-domain QoS routing schemes to
able to manage the LSDBs effectively and to minimize a
waste of the network system resources by LSU massages. In
Inter-Domain QoS Routing Scheme using Link State Information

Tae-IL Kim*, Hae-Won Jung*, Min Young Chung**, and Seong-IL Jin***

Electronics and Telecommunication Research Institute*
161 Gajeong-Dong, Yuseong-Gu, Daejeon, 305-500, Korea
School of Information and Communication Engineering, Sungkyunkwan University**
300 Chunchung-Dong, Jangan-Gu, Suwon, 440-746, Korea
Department of Computer Science, ChungNam National University***
220 Kung-Dong, Yuseong-Gu, Daejeon, 305-764, Korea
{tikim, hw-jung}@etri.re.kr*, mychung@ece.skku.ac.kr**, sijin@cs.chungnam.ac.kr***

ISBN 978-89-5519-136-3-1254- Feb. 17-20, 2008 ICACT 2008

Page 2

Section II we propose the inter-domain QoS routing scheme
using link state information. Section III introduces the basic
scheme, the conventional scheme, and describes the network
topologies for simulation. In Section IV, we evaluates the
performance of proposed scheme compared with the basic
scheme and the conventional scheme. Finally, we make
conclusions in Section V.

II. The Proposed Schemes for Inter-Domain QoS
Routing
If source and destination routers are in different domains,
a service flow passes through gateway routers. Therefore, the
information of the link connected with gateway routers is
more important than the other link state information because
the service flow necessarily passes through the link of the
both gateway routers connecting two different domains.
There are two important links connected to the gateway
router, the external links and the internal links. The external
link is the links connecting different domains, that is, BGP
peer link. The internal link is all links except the external link.
We propose that a router manages one of the two above
link state information and sets up a QoS path using the
information. There are two schemes according to the
information that routers manage. In the proposed scheme 1,
routers manage the mean available bandwidth about internal
link of a gateway router. On the other hand, the proposed
scheme 2 is that routers manage available bandwidth of
external link of gateway router.
In both proposed schemes, gateway routers transmit
important link information when they transmit network
reachability information to inner routers. Therefore every
router manages reachability and important link information,
which is called 'representative value'. When a router receives
a service request, the router uses the representative value for
QoS path. The following explains the process for setting up
QoS path using the proposed scheme.
1) A source router receiving a request for QoS path
compares the representative values about the
destination router and sends request message for
setting up QoS path to a gateway router which has
the largest representative value.
2) The gateway router which receives the request selects
a external link based on representative value and
requests the gateway router of next domain to set up
QoS path.
3) The gateway router of next domain calculates QoS
path. If the destination and gateway router are in
same domain, the gateway router sets up a QoS path
to the destination router based on LSDBs. If not, the
gateway router sets up QoS path to the next gateway
router by the procedure of 1).
4) Until QoS path is set up to the destination router, the
procedure 2) and 3) are repeated.
The update of the important link state information is
performed by the gateway router. The gateway router updates
the representative value based on LSU algorithm.
Figure 1 illustrates an example that there are two paths
from a source router to a destination router. In this example, a
source router is R1 and a destination is R2. When two
proposed schemes are applied, a gateway router manages
important link (internal link or external link) information,
respectively. If representative value of G1 and G5 are 3Mbps
and 2Mbps, the gateway routers, G1 and G5 that belong to
AS1, transmit reachability information and representative
values to all routers inside AS1 in forms of (G1, R2, 3Mbps)
and (G5, R2, 2Mbps). In this case, R1 selects the path that
passes through G1 because the representative value of G1 is
larger than that of G5.
AS1 AS1
AS3 AS3
AS2AS2
G1G1
G2 G2
G3 G3
G4G4
G5G5
G6G6
R1R1
R2R2

Figure 1. An example of inter-domain QoS routing

III. Network Topology for Simulation
The current Internet based on IPv4 consists of Stub
Domain, Transit Domain and Stub-stub edge. Stub domain
only carries traffic which is generated or terminated in a sub-
network such as campus networks or collections of local area
networks (LANs). Multi-homed stub domain has connections
to more than one Transit Domain. Transit Domain is defined
as a set of backbone nodes and links, and it is connected to a
number of Stub Domains. Stub Domain can also be
connected directly through Stub-stub edges, not through
ISBN 978-89-5519-136-3 -1255-Feb. 17-20, 2008 ICACT 2008

Page 3

Transit Domain[7].
To evaluate the performance of the proposed schemes,
we define a network topology as shown in Figure 2. We
implement traffic of Stub Domain by generating traffic from
backbone node of Transit Domain and do not implement
nodes inside Stub-domain individually. We suppose that
Multi-homed stub domain and Stub-stub edge is not exist.
And we assume that the connection requests arrive as a
Poisson process and the holding duration of a service is
determined by an exponential distribution. The request
bandwidth is uniformly chosen in the range of [1Mbps,
5Mbps] and range of [3Mbps, 7Mbps]. Every time the link
state changes all routers update LSDBs and we do not
consider propagation delay.
0
1
2
34
5
6
7
8
9
11
10
12
13
14
15
16
17
18
19
20
21
22
23
25
24
26
27
42
43
44
45 46
47
48
49
50
51
53
52
54
55
28
29
30
31
32
33
34
35
36
37
39
38
40
41
Transit Domain 0Transit Domain 1
Transit Domain 3Transit Domain 2

Figure 2. Network topology for simulation
We introduce the basic scheme and the conventional
scheme in order to compare the performance of the proposed
schemes. The basic scheme is the ideal case that all routers
know link state information of the whole topology. In this
case, a source router sets up the inter-domain QoS path in the
same way as intra-domain QoS routing. Therefore, the router
can accurately set up a QoS path[8]. On the other hand, in the
conventional scheme, all routers only know link state
information about its own domain. Since a source router
cannot know network topologies or link state information of
other domains, the source router set up a QoS path via a
default gateway router[9]. If there is not a default gateway
router, a protocol selecting a gateway router will be needed.
In this simulation, we assume that there is not a default
gateway router. Therefore, a source router excludes the links
that have a smaller bandwidth than the request bandwidth and
executes the shortest path algorithm with gateway routers
which provide reachability information. After that, the source
router chooses the nearest gateway router as a middle router
to pass by.
In order to measure the performance, we define the
number of updates per unit time and blocking probability.
The number of updates per unit time is defined as the total
number of transmitted LSU messages divided by the total
simulation time. The block of connection request is divided
into Type 1 and 2. Type 1 block occurs when there is not a
path from source to destination router after calculating the
path based on LSDBs. When a path from source to
destination router exists in LSDBs but it is impossible to
provide the service in real link, Type 2 block occurs. The
total number of block is the sum of the number of blocked
connections with Type 1 and Type 2.

IV. Performance Evaluation
Figure 3 shows the block probability in the network
topology. In the basic scheme, the blocking probability is the
smallest since all routers know link state information of the
total topology. On the other hand, the blocking probability of
the conventional scheme is larger than other schemes because
the source router does not know link state information of an
external domain. The proposed schemes have lower blocking
probability than the conventional scheme because routers
know summarized link state information of other domains.
0 50 100 150200250300350400450500
1E-4
1E-3
0.01
0.1
block probability
Total offered traffic load (arrival rate/service rate)
Basic scheme
Conventional scheme
Proposed scheme

Figure 3. Block probability varying traffic load(Request
bandwidth is range of [1Mbps, 5Mbps])
There is a trade-off between blocking probability and
link state update rate. In basic scheme, although blocking
probability is the small, routers update link state information
ISBN 978-89-5519-136-3-1256- Feb. 17-20, 2008 ICACT 2008

Page 4

frequently as shown in Figure 4. Conventional scheme has
low link state update rate, but this scheme shows high
blocking probability. The
comparatively low link state update rate and has smaller
blocking probability than the conventional scheme.
proposed schemes have
0 50100150200250300350400450 500
0
50000
100000
150000
200000
250000
300000
350000
400000
Basic scheme
Conventional scheme
Proposed scheme
Number of updates per unit time
Total offered traffic load (arrival rate/service rate)

Figure 4. Link state update rate varying traffic load(Request
bandwidth is range of [1Mbps, 5Mbps])
To evaluate the performance of proposed scheme in
different request bandwidth, we change the range of request
bandwidth from [1Mbps, 5Mbps] to [3Mbps, 7Mbps]. Figure
5 shows the blocking probability. As the request bandwidth
increases, the blocking probability increases. However, the
trend of blocking probability is similar as shown in Figure 3.
0 50 100150200250300350 400450500
1E-4
1E-3
0.01
0.1
block probability
Total offered traffic load (arrival rate/service rate)
Basic scheme
Conventional scheme
Proposed scheme

Figure 5. Block probability varying traffic load(Request
bandwidth is range of [3Mbps, 7Mbps])
0 50 100150200250300350 400450500
0
50000
100000
150000
200000
250000
300000
Basic scheme
Conventional scheme
Proposed scheme
Number of updates per unit time
Total offered traffic load (arrival rate/service rate)
Figure 6. Link state update rate varying traffic load(Request
bandwidth is range of [3Mbps, 7Mbps])
Figure 6 illustrates the number of updates per unit time
when the request bandwidth is range of [3Mbps, 7Mbps] in
the network topology. In this case, the link state update rate is
lower than in the range of [1Mbps, 5Mbps]. If request
bandwidth increases, blocking probability increases and the
number of acceptable service is decreases. Since the link state
changes small when services are accepted a little, link state
update rate decreases as request bandwidth increases.
However, the trend is similar to the results in Figure 4.

V. Conclusion
In inter-domain QoS routing, it is unrealistic for a router
to manage entire link state information on all domains. In this
paper, we proposed the inter-domain QoS routing schemes
that can manage the LSDBs effectively and minimize a waste
of the network system resources by LSU massages. The
performance of the proposed scheme was evaluated by
simulations. The performance of two proposed schemes
showed better than conventional scheme in view of blocking
probability. It is also better than basic scheme in view of link
state update rate. The link state update rate of the proposed
schemes is not more sensitive than basic scheme to the traffic
load. Therefore the proposed schemes can reduce the
blocking probability and control the link state update rate. In
addition, it is possible to embody the proposed schemes into
internet routers because a router manages only a few link
state information.

ISBN 978-89-5519-136-3-1257- Feb. 17-20, 2008 ICACT 2008

Page 5

References
[1] F. A. Kuipers and P. F. A. Van Mieghem, “ Conditions That
Impact the Complexity of QoS Routing,” IEEE/ACM
Transactions on Networking, Vol. 13, No. 4, August 2005,
pp.717-730.
[2] G. Apostolopoulos, R. Guerin, S. Kamat, A. Orda, and S. K.
Tripathi, “ Intradomain QoS Routing in IP Networks: A
Feasibility and Cost/Benefit Analysis,” IEEE Network, Vol. 13,
September/October 1999, pp.42-54.
[3] L. Xiao, K.-S. Lui, J. Wang, and K. Nahrstedt, “ QoS
Extension to BGP,” Proc. of IEEE ICNP 2002, November
2002, pp.100-109.
[4] L. Xiao, J. Wang, K.-S. Lui, and K. Nahrstedt, “ Advertising
Interdomain QoS Routing Information,” IEEE Journal on
Selected Areas in Communications, Vol. 22, Issue 10,
December 2004, pp.1949-1964.
[5] L. Benmohamed and B. Doshi, “ QoS Routing in Multi-
Domain Networks,” Proc. of IEEE Pacific Rim Conference on
Communications, Computers and Signal Processing, August
2005, pp.137-140.
[6] S. Norden and J. Turner, “ Inter-Domain QoS Routing
Algorithms,” Department of Computer Science Technical
Report, WUCS-02-03, 2002.
[7] K. I. Calvert, M. B. Doar, and E. W. Zegura, “ Modeling
Internet Topology,” IEEE Communications Magazine, Vol. 35,
No. 6, June 1997, pp.160-163.
[8] M. Shankar, M. D. Miguel, and J. W. S. Liu, “ An End-to-End
QoS Management Architecture,” Proc. of RTAS 1999, June
1999, pp.176-189.
[9] Y. Rekhter and T. Li, “ A Border Gateway Protocol 4 (BGP-
4),” IETF RFC 1771, March 1995.

ISBN 978-89-5519-136-3-1258-Feb. 17-20, 2008 ICACT 2008