INTERNET PROTOCOL / MULTI-PROTOCOL

LABEL SWITCHING

Owakah Nwobike Innocent

Abstract—MPLS (Multi-protocol Label Switching) technology within IP networks, most of them may end up consuming
emerged and became popular recently with the advantage of fast more system resource than expected because of the
forwarding. And also it has brought a great opportunity to both dramatically high data rates in nowadays highly burdened
Internet Providers and users by giving a much faster data Internet, especially in the core area where most of the data
exchanging environment into almost every aspect of the Internet, exchanging operations took place. MPLS technology itself
such as unicasting, multicasting, VPN(Virtual Private Network)
needs to be optimized before it could be added into IP
service and even Traffic Engineering with the help of RSVP
(Resource reSerVation Protocol) in both IPv4 and IPv6 networks.
networks, or else it won’t last long before another bottleneck
Much work has been done by researchers all over the world to hits the Internet again.
give suggestions about how to better integrate MPLS technology
II. BACKGROUNG KNOWLEDGE
into IP networks. Unfortunately, much less needed progress has
been made on MPLS technology itself in these years, which led to A. Multi-Protocol Label Switching
bottlenecks in delivering high speed and QoS (Quality of Service)
guaranteed services in future NGN (Next Generation Network) Multi-Protocol Label Switching was originally introduced in
environment. In this article, we proposed a layered architecture 1970s by several Internet engineers with the initial purpose of
for MPLS network itself. Based on the fact that MPLS technology getting rid of the drawbacks in packet switching mechanism,
is independent from upper layer network protocols, this proposed in which every packet’s header should be read and analyzed
mechanism could be easily integrated into both IPv4 networks before sending it out for the right destination.
and future IPv6 networks, and help to upgrade the whole After almost 20 years of struggling and developing, MPLS
performance of the Internet environment. was finally received and confirmed by the IETF (Internet
Engineering Task Force) in late 1990s. This moment was also
considered to be the start point that MPLS technology became
Keywords-words; MPLS; RSVP; Traffic Engineering; QoS the open standard to be ready for adoption [3].
MPLS technology is actually performed between the L2
I. INTRODUCTION and L3 network, which makes it to be called the L2.5
As many high-bandwidth consuming Internet services, such as technology. The idea of using short labels (20 bits) rather than
longer IP addresses (32 bits in IPv4 and 128 bits in IPv6) in it
net-conferencing, Video on Demand, Internet games and so comes from the Virtual Circuits used in ATM (Asynchronous
on, start to dominate the Internet world, both Internet users Transfer Mode) technology. Figure 1 gave a snip pick on the
and providers are facing the dilemma between limited structure of an MPLS packet header [4].
bandwidth and boosting requirement for Internet accessing
speed. In order to overcome this problem, MPLS [1][2].
Technology has been recently introduced to supply higher data
exchanging speed.
Since it’s born in late 1990s, MPLS technology has been
used by many global Internet Service Providers in building up Figure 1. MPLS packets header structure
new service structures and delivering customer services.
Besides, many research centers have also put lots of time and There are 4 segments in an MPLS header: Label, EXP, S and
efforts in studying its best usage in both IPv4 and IPv6 TTL. Label contains the value of actual label added to the data
networks, and given their own suggestions on this subject. packets; EXP is often preserved for special usage, such as in
Though there are many ways of binding MPLS technology QoS operation; S is the indicator for bottom label; and TTL is
short for Time to Live, shows how far the header could travel
along the route. mechanism in MPLS, data transportation speed boosts
effectively comparing with packet switching networks. But
still, there are some issues in this basic MPLS network
III. ROUNTING MACHANISM OF MPLS architecture need to be solved:
The mechanism of using labels gives MPLS technology the 1. As the network itself and number of EUs keep grows,
ability of transferring data in a very high speed. The details of more and more labels would be required in order to meet the
data transferring procedure in MPLS networks are introduced users’ requirement. If this continues, the FEC database on
below with a basic MPLS network architecture in Figure 2. every PE router will become bigger and data transferring will
consume more network resource, which in turn may drag
down the performance of the whole system.
2. Even EXP segment in MPLS header makes it possible
to achieve QoS by integrating L2 QoS method, such as
DiffServ and IntServ architecture. But that’s not enough for
the future if MPLS technology wants to be in the NGN
environment [7].
These two major issues mentioned above should be
Figure 2. Architecture of basic MPLS network [5]. carefully and thoroughly fixed before MPLS technology could
be further considered as the better choice rather than packet
As shown in Figure 2, there are three different kinds of routers switching methods.
in MPLS network architecture: PE routers, P routers and CE
routers. The former two kinds are MPLS enabled, while CE is IV MANAGEMENT OF MPLS
just a regular accessing router. P router is used in the core area
doing nothing but forwarding labeled data packets; PE router The Management based concepts in managing an MPLS
is used for data encapsulation and decapsulation and bridging network, because it is considered as an appropriate way of
the P routers and CE routers. There are two sets of routing dealing with large sets of managed elements. Using policy-
protocols running independently on PE routers. One of them is based management for networks and systems has become very
running between PE routers and CE routers changing IP and popular since the early work on policies such as [8].
connection information, while the other is used between P Nowadays, some commercial products are available, which
routers and PE routers for label distribution and LSPs (Label use some form of policies to configure and control networks.
Distribution Paths) management [5]. In the IETF there is a Policy Framework Working Group [9],
Supposing in Figure 2, EU1 (short for End User) in Access which aims at resolving issues related to policy-driven
Network I wants to send data to EU2 in Access Network II, management of IP networks. It includes the definition of a
the data forwarding procedure in this basic MPLS network policy framework and information models for DiffServ,
architecture is described as follow: IntServ, and IP Devices. The IETF policy framework activities
are on one hand limited to DiffServ/IntServ based networks,
First, IP packets will be sent to CE1 for further forwarding. and on the other mainly dealing with device configuration.
CE1 will check the destination address and look up its routing
table to find out if the data packets should go up to PE1 or • A Policy Definition Language (PDL) is used to define
down to other EUs. new policies in terms of policy rules with condition and
action lists. What language to use is very controversial,
Packets are sent up to PE1, where the destination address and the IETF has not reached consensus in
will be checked again and matching labels will be chosen and standardizing a Policy Definition Languages. Basically,
added onto the packets according to PE1’s FEC (Forwarding each implementation defines its own language.
Equivalence Class) database, which is used in every PE router
for storing the mapping relationship between outgoing labels • The simple version of the X.500 directory access
and certain destination addresses. After encapsulation, labeled protocol called Light-Weight Directory Access Protocol
data packets will be sent to P routers for fast forwarding in (LDAP) is used by the policy server to retrieve
core area. information from the repository. Note that any other
database may be used, but the working group decided to
Every hop for a data packet in the MPLS area should be only provide a mapping of the policy model to a LDAP
companied with label switching, which means pushing out old schema.

 
labels and popping in new labels [6]. Management
Console
After traveling along the predefined route and going
through core area, the labeled packets will arrive at PE2, get
decapsulated (strip off the labels) and forward to CE2. PDL
Policy Decision Engine
CE2 would check the destination address and send the data (PDE) Network Information Policy Server
packet to EU2. Once EU2 gets the packet, this whole Model Policy Information
procedure is finished. By using this label switching

PE PE
PEC
 Figure 3: IETF Policy Framework [4]. The native service provided to end users is the IP service. This
is a Layer 2-based IP service that is distinguished from a
One of the key issues in the framework is how policy rules
are triggered by state transitions or events. A Policy Layer 3 IP service by the fact that a PE forwards a CE’s IP
Decision Engine (PDE) is typically used to handle requests. traffic based upon Layer 2 information rather than Layer 3
For instance, in COPS for RSVP, the PEP issues a COPS information. In particular, the PE device does not perform
request and the policy server returns a decision on whether longest prefix match lookup of the destination IP address for
to permit or deny the RSVP Path Message. A table of frame forwarding, nor does it participate in routing protocols
enabled policy rules is traversed at the PDE in order to find with the CE, for discovering network topology. In the context
the matching rules for a request. The scenario is fairly clear of Virtual Private Networks, this Layer 2 forwarding
for QoS signaling using the Resource Reservation Protocol characteristic makes this a Layer 2 VPN service as opposed to
(RSVP). RSVP path and reservation messages arriving at an an IP-VPN (L3VPN) service [15].
RSVP daemon running on an IP router are converted by a The use of Layer 2 interworking for IP traffic enables carriers
COPS client component into COPS requests and sent to the and service providers to introduce Ethernet and PPP as
COPS server component at the policy server. The RSVP- attachment while preserving existing ATM and frame Relay
related COPS request will be forwarded to the decision infrastructures.
engine which makes a decision based on the applicable
rules [10][11].
VI QoS in IP / MPLS
V IP OVER MPLS
In spite of their numerous practical advantages, the Next
Generation Network (NGN) is not without limitations. The
main deficiency of the modern NGN model lies in the fact that
it may not sufficiently guarantee an end-to-end QoS for multi-
service traffics. Consequently, current NGN networks may
require more advanced control mechanisms in the control and
transport levels to overcome some of these deficiencies. In
particular, the QoS signaling and control structure of the
current NGN model is the aspect that requires improvement to
harmonize the QoS needs with the network capacity. In this
work, we propose an alternative NGN model which provides
the desired improvement to satisfy the QoS requirements.

Figure 4: IP over MPLS Structure [4]. Currently, the QoS signaling and the NGN networks control
solutions are still at the development stages. Packet Cable,
which supports QoS in packet-based cable access networks for
IP Multi-service Interworking over MPLS is a Virtual Private telephony services, proposed a QoS solution [16]. This
Wire Service (VPWS). It provides point-to-point IP solution focuses mainly on some specific problems related to
connectivity for two CE devices across an MPLS network. packet-based cable access networks. Similarly, the 3GPP also
What distinguishes this from a homogeneous Layer 2 VPN is proposed an end-to-end QoS solution for the 3rd generation
that it allows the type of Attachment Circuits used to connect mobile networks [17]. It was developed for specific category
each site to be different from each other, and it does so in a of networks and lacks several functionalities that can be
way that requires minimal configuration changes to the deemed necessary for a standard NGN QoS model (core
connecting CEs in some topologies and no changes in other network resource control, topology knowledge, path selection,
topologies [12]. etc.). Also, the choice of Diameter protocol for Gq interface
This describes how a PE device discovers a CE device that is must be more argued for what this protocol is generally used
attached by one of a variety of types of Attachment Circuit, to perform resource reservation requests.
and how a PE forwards a customer’s IP traffic based upon the
learned Layer 2-specific information. The solution requires the VII Conclusion
PE to participate in the different address resolution procedures
used by CE devices on different types of Attachment Circuit, This paper has scrutinized the common assumption that MPLS
and to exchange the information with other PEs to enable a reduces packet forwarding time with respect to IP due to the
local PE to act as a proxy for a remote CE [13]. fast ILM look-up. Therefore, an extensive performance
This specification describes Multi-service Interworking for IP evaluation has been performed using Intel’s IXP1200 network
version 4 over an MPLS core network. The term "Multi- processor and no specialized hardware such as a Content
service Interworking" means that IP packets are transported Addressable Memory or search engines. On the network
across the MPLS core in a pseudo wire (PW) , and the two processor the basic functionality of ingress LER, an LSR and a
attachment circuits (ACs) associated with the PW may employ proxy egress LER as well as that of an IP router has been
different Layer 2 technologies (e.g. ATM on one AC and implemented.
Ethernet on other) [14].
On the basis of that implementation, it has been shown that the [7] F. Blanchy, L. MClon, and G. Leduc. An efficient decentralized on-line
tra51c engineering algorithm for MPLS networks. Available at
delay at an LSR is sometimes smaller and sometimes larger http://run.montefiore.ulg.ac.be/blanchy,july 2002.
compared to that at an IP router, which depends primarily on [8] M. Masullo, S. Calo, “Policy Management: An
the packet size and the transmission rate at the outgoing Architecture and Approach”, Proceedings of the First
interface. The main conclusion is that the common assumption IEEE Intl. Workshop on System Management, LA, USA,
that MPLS generally allows for faster packet forwarding than April 1993.
IP does not even hold on standard hardware. However, it may [9] J. Moffett, M. Sloman, ―Policy Hierarchies for
be concluded with a certain degree of approximation that an Distributed Systems Manage-ment‖, IEEE JSAC Special
LSR is faster for smaller packets of less than about 150 Byte Issue on Network Management, Vol 11, December 1994.
(IP packet size) and about as fast as IP for packets of more [10] R. Wies, ―Policies in Network and Systems Management -
Formal Definition and Architecture‖, Journal of
than about 150 bytes.
Networks and Systems Management, Vol 2(1), 1994.
[11] IETF Policy Framework Working Group,
REFERENCES http://www.ietf.org/html.charters/ policy-charter.html.
[1] E.Rosen, A.Viswanathan, R.Callon. Multiprotocol Label [12] Li, T., Rekhter, Y. " A Provider Architecture for
Switching Architecture[S], RFC 3031. January 2001, Differentiated Services and Traffic,
[2] E.Rosen, G.Fedorkow, Y.Rekhter, D.Farinacci, T.Li, Engineering(PASTE)", RFC 2430, October 1998.
A.Conta. MPLS Label Stack Encoding[S], RFC 3032. [13] Awduche, D.O., Malcolm, I., O'DELL, M., McMmus, J.
January 2001. "Requirements for Traffic draft-ietf-mpls-traffieeng-
[3] E.Rosen et al, ‖Multiprotocol Label Switching OO.txt, OCTOBER 1998 Engineering over MPLS".
Architecture‖, RFC 3031, January 2001. [14] Rosen, E.C., Vishwanathan, A., Callon, R. "Multiprotocol
[4] B .Davie et al, ― MPLS using LPD and ATM VC Label Switching Architecture", dmft-ietf-mpls-d44.txt,
Switching‖, RFC 3035, January 2001. Feb. 1999.
[5] Murali S. Kodialam and T. V. Lakshman. Minimum [15] Eric c. Rosa, Y. Rekhter, "MPLS Label Stack Encoding",
interference routing with applications to MPLS traffic Drafting –ietf-mpls-label-encaps-03.txt, Sept 1998.
engineering. In ZNFOCOM (2), pages 884493,2000.
P. Aukia, M. Kcdialam, P. Koppol, T. Lakshman, H. [16] PacketCable 1.5 Specifications PK-TSP-DQOS1.5-I02-
[6]
Sarin, and B. Suter. RATES: A server for MPLS traffic 050812, Dynamic Quality –of -Services" PacketCable,
engineering. RATES: A server for MPLS traffic August 2005.
engineering, IEEE Network Magazine, pp. 3441, [17] 3GPP TS 23.207, "3rd Generation Partnership Project;
MarchlApril2000. Technical Specification Group Services and System
Aspects; End-to-end Quality of Service (QoS) concept
and architecture‖ 3GPP, September 2004.