International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-2, Issue-3, March 2014
Futureproofing an Enterprise Network with
MPLS/BGP
Tina Satra, Smita Jangale
Border Gateway Protocol (BGP), which acts as a core routing
protocol of the Internet. It maintains a table of IP networks or
Abstract Enterprise used VSAT(Very Small Aperture prefixes and assigns network reachability to autonomous
Terminal), a satellite communication system for data transfer. systems; as a result it is indirectly used by the Internet users.
It had many drawbacks like small data transfer window, no iBGP protocol is used among the routers in autonomous
guarantee over analog transmission, low bandwidth, no system to command the internal routers.
security, less cost effective, etc. Then private networks, based
on Frame Relay and point-to-point circuits where used which
provided network security since public did not have access.
II. LITERATURE REVIEW
Later Frame Relay with Permanent Virtual Circuits (PVCs) Multiprotocol Label Switching is a popular networking
were used as Virtual Private Network (VPN), where private technology that uses labels attached to packets to forward
data was transferred over public networks, since enterprises them through the network. Before MPLS, the most popular
needed dedicated PVCs cost was more. So, emerging category WAN protocols were ATM and Frame Relay, then, with
now is Service Provider VPNs, where a service provider
popularity of the Internet, IP became the most popular. The
provides the backbone network to an enterprise. Service
Providers can use MPLS/BGP approach, where MPLS MPLS labels are advertised between routers so they can build
(Multiprotocol Label Switching) forwards the data supporting a label-to-label mapping. These labels are attached to the IP
multiple protocols and BGP (Border Gateway Protocol) packets, enabling the routers to forward the traffic by looking
controls route to construct secure network [1], [2]. The paper at the label and not the destination IP address. Thus, the
depicts simulation of MPLS/BGP network for enterprises, packets are forwarded by label switching instead of IP
using GNS3 simulator. Two enterprises working on Open switching. This has lead to popularity of MPLS. BGP/MPLS
Shortest Path First (OSPF) and Enhanced Interior Gateway VPN supports the provision of IP connectivity by a service
Routing Protocol (EIGRP) having sites at different location are provider to multiple customers over a common physical IP
depicted. Virtual routing and forwarding instant, working on
backbone, while allowing complete logical separation of
label switching and node failure (redundancy) for this network
is discussed. How the provider network restricts access to customer traffic and routing information. Interconnection of
different enterprises is shown. The performance of MPLS different sites belonging to the same customer is provided
exceeds conventional IP routing is depicted using test cases. over the MPLS backbone. Fig. 1 shows an example of a VPN
The results show that on node failure alternate path change with a service provider (P) backbone network, service
takes very few seconds. provider edge routers (PE), and customer edge routers (CE).
The key network components of the BGP/MPLS VPN are the
provider edge (PE) routers, the provider (P) routers and the
Index Terms MPLS, Enterprise Network, BGP-MPLS, customer edge (CE) routers. The PE routers are routers
MPLS Performance. within the service provider backbone that connect to
customer sites. In a MPLS network, a PE router also
I. INTRODUCTION performs as an edge LSR. The P routers are routers within the
To integrate the work in an enterprise, employees share service provider backbone that do not connect directly to
files on the networking system. Today, enterprises use customer sites. They are the LSRs in a MPLS network. The
applications like customer relationship management (CRM), CE routers are routers at the customer sites that are directly
enterprise resource planning (ERP), etc., commonly. So connected to the service provider network. They connect
servers hosting these applications have to now process and directly to the PE routers.
deliver data at Gigabit speeds. These performance demands
and increase in bandwidth must be supported by the III. NETWORK SIMULATION
enterprises network infrastructure. Thus, a secure a reliable Network Structure for Simulation is depicted in Fig. 1. There
network connection is required to protect the data transferred are two Provider routers P1 and P2, four Provider Edge
[3]. Multiprotocol Label Switching (MPLS) offers extremely routers PE1, PE2, PE3 and PE4 and two Customer Edge
scalable, deterministic re-route, traffic engineering, routers CE1A and CE1B. Customer IBM-A site is connected
data-carrying mechanism which transfers data packets with to CE1A with and customer IBM-B site is connected to CE1B
assigned labels across the network through virtual links [5]. with. OSPF 1 area 0 is given to all P and PE routers while
OSPF 10 area 1 is given to CE1A and CE1B routers. Since
Manuscript received March 01, 2014. MPLS supports different protocols, simulation is extended to
Tina Satra, Department of Computer Engineering, Shah and Anchor
Kutchhi Engineering College, Mumbai, India, +91 08454040291.
Customer APL. Customer Edge routers CE2A and CE2B are
Smita Jangale, Department of Information Technology, VES Institute of extended to PE2 and PE3 respectively, and run EIGRP
Technology, Mumbai, India, +91 09869190837
37 [Link]
Futureproofing an Enterprise Network with MPLS/BGP
process 10. Customer APL-A site is connected to CE2B and packet with label 27 would remove the label and perform an
customer APL-B site is connected to CE2A. IP lookup on the destination IP address.
Fig. 3 Label Switching on routing path
C. Node Failover (Redundancy)
The MPLS network designed is such that redundancy exists.
There are multiple paths to reach destination address. When
a link fails, automatically other path is taken to reach the
destination address. Fig. 4a and fig. 4b shows alternate path
Fig. 1 Network Structure. taken to reach IBM-B site. When P1 fails P2 ([Link]) is
chosen to complete the route. Figure 3.6.2 shows that packet
A. Virtual Routing and Forwarding Instant
destined for [Link] entering the MPLS network on ingress
How an IP packet traverses the MPLS VPN backbone from
LSR (PE1), where it is imposed with the label 24 and
one customer site to another? MP-BGP running between the
switched towards the next LSR. Second LSR (P1) swaps
PE routers distributes the vpnv4 routes and their associated
incoming label 24 to outgoing label 27 and forwards to next
VPN label. Between all PE and CE routers, the routing
LSR (PE4). The egress LSR receiving a packet with label 27
protocol puts the customer routes into the VRF routing table
would remove the label and perform an IP lookup on the
on the PE routers. Fig. 2 shows path taken by the packet from
destination IP address.
CE1A, where IBM site A is located, to CE1B, where IBM site
B is located. The packet routes through CE1A([Link]) -
PE1([Link]) P1([Link]) - PE4([Link]) CE1B
([Link]).
Fig. 4a Alternate route to IBM-B site.
Fig. 2 Path from IBM-A site to IBM-B site.
B. Working of Label Switching
Fig. 3 shows that packet destined for [Link] entering the
MPLS network on ingress LSR (PE1), where it is imposed
with the label 21 and switched towards the next LSR. Second
LSR (P1) swaps incoming label 21 to outgoing label 27 and
forwards to next LSR (PE4). The egress LSR receiving a
38
International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-2, Issue-3, March 2014
Fig. 4b Alternate route to IBM-B site.
Limited capability to deal with addressing
IV. TEST CASES AND RESULTS information.
All traffic to same IP address is treated similar.
A. Restricts others in Enterprise VPN It becomes difficult to perform traffic engineering.
One of the MPLS traffic engineering is to restrict admission For highly interactive application, flow of packets
to other Enterprise VPNs. Thus if APL site tries to contact should have low delay and less packet loss
IBM site, MPLS will restrict its admission. Fig. 5 shows that threshold.
APL-A site tries to contact IBM-A site or IBM-B, it fails to Routing in conventional IP networks take excess time due to
so. Similarly, IBM site is restricted to access APL VPNs. look up tables. Also re route on node failure is difficult in
Fig. 6 shows that IBM-A site is not allowed to contact APL conventional IP routing. MPLS allows fast re route
sites. capability due to label switching. To check the performance
of MPLS network on node failure was performed with
following test cases. Time taken to re-route is calculated
using time out for packet to reach destination with number of
dropped packets before new route is selected. Six test cases
were used to find average time taken to re-route on node
failure in the network. Test case 1 and test case 2 were taken
for packet flow from IBM-A site to IBM-B site, where
packets take route from provider P1. When P1 fails, packets
drop until MPLS re-routes packet from P2. The scenarios are
shown in fig. 7 and fig. 8. In test case 1, 13 packets are
dropped before re-route and in test case 2, 12 packets are
dropped before re-route.
Fig. 5 Restriction to APL site on contacting IBM site.
Fig. 7 Test case 1 (re-route from IBM-A site to IBM-B site).
Fig. 6 Restriction to APL site on contacting IBM site.
B. Performance test
In IP routing, conventionally, each router has to
independently decide route for each incoming packets. The
router consults the routing table, which is build using IP
routing protocols like BGP, OSPF, IS-IS, etc., to find the next
hop for that packet based on destination address. Each router
in the network performs all of these steps. The main issue
with conventional routing protocols is that they do not take
capacity constraints and traffic characteristics into account
when routing decisions are made. There are many limitations
in conventional IP routing.
Fig. 8 Test case 2 (re-route from IBM-A site to IBM-B site).
39 [Link]
Futureproofing an Enterprise Network with MPLS/BGP
Fig. 9 Test case 3 (re-route from APL-B site to APL-A site).
Test case 3 and test case 4 were taken for packet flow from
APL-B site to APL-A site, where packets take route from
provider P2. When P2 fails, packets drop until MPLS
re-routes packet from P1. The scenarios are shown in fig. 9
and fig. 10. In test case 3, 9 packets are dropped before
re-route and in test case 4, 11 packets are dropped before Fig. 12 Test case 6 (re-route from APL-A site to APL-B site).
re-route.
Test case 6 was taken for packet flow from APL-A site to
APL-B site, where packets take route from provider P2.
When P2 fails, packets drop until MPLS re-routes packet
from P1. The scenario is shown in fig. 12. In test case 6, 10
packets are dropped before re-route.
Based on test cases following data is computed.
Re-route time = No. of dropped packets X time out.
Table 1 gives average re-route time on node failure. Figure
14 shows bar graph for test cases depicting re-route time.
Thus, from above given data we conclude that average
re-route time on node failure in 21.67 seconds, which is very
less than time taken to re-route in IP routing using fast
re-route techniques.
Fig. 10 Test case 4 (re-route from APL-B site to APL-A site. V. CONCLUSION
In traditional IP routing, there are many limitations. Like
Test case 5 was taken for packet flow from IBM-B site to limited capability to deal with addressing information. Re
IBM-A site, where packets take route from provider P2. route mechanism not available instantly. All traffic to same
When P2 fails, packets drop until MPLS re-routes packet IP address is treated similar. Routing packets in a network
from P1. The scenario is shown in fig. 11. In test case 5, 10 with two same IP addresses of different customers cannot be
packets are dropped before re-route. handled. Highly interactive applications in enterprise cannot
run smoothly with such routing of packets. In conventional
IP routing using fast re-routing techniques, time taken to
re-route requires recalculating routes, which may go up to
thousands of seconds.
In BGP-MPLS VPN, time taken to re-route on single node
failure is just few seconds. Handling route for same IP
address for different customer is not an issue since routing is
done using labels and not IP addresses. Customers are not
aware of MPLS and how packet route to destination.
Different customers are restricted access of each others data,
thus privacy is maintained. Since MPLS is supporting
multiple protocols, OSPF and EIGRP VPNs could
Fig. 11 Test case 5 (re-route from IBM-B site to IBM-A site).
40
International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-2, Issue-3, March 2014
successfully utilize same network to route packet, thus
reducing cost to enterprise.
Table 1. Average time taken to re-route on Node failure.
Test Case Test Case (packets send Actual Path taken No. of Time taken New Path taken
No. from to customer sites) dropped to re-route
packets (seconds)
1 IBM-A to IBM-B CE1A-PE1-P1-PE4-CE1B 13 26 CE1A-PE1-P2-PE4-CE
1B
2 IBM-A to IBM-B CE1A-PE1-P1-PE4-CE1B 12 24 CE1A-PE1-P2-PE4-CE
1B
3 APL-B to APL-A CE2A-PE2-P2-PE3-CE2B 9 18 CE2A-PE2-P1-PE3-CE
2B
4 APL-B to APL-A CE2A-PE2-P2-PE3-CE2B 11 22 CE2A-PE2-P1-PE3-CE
2B
5 IBM-B to IBM-A CE1B-PE4-P2-PE1-CE1A 10 20 CE1B-PE4-P1-PE1-CE
1A
6 APL-A to APL-B CE2B-PE2-P2-PE3-CE2A 10 20 CE2B-PE2-P1-PE3-CE
2A
Average 10.83 21.67
REFERENCES
[1] E. Rosen, Y. Rekhter, BGP/MPLS IP Virtual Private Networks (VPNs)
RFC 4364, February 2006.
[2] E. Rosen, Applicability Statement for BGP/MPLS IP Virtual Private
Networks (VPNs) RFC 4365, February 2006.
[3] Alcatel, Lucent, MPLS-Enabled Network Infrastructures 2007.
[4] ADTRAN, Inc. Private IP Service BGP/MPLS VPN Networks 2005.
[5] De Ghein, MPLS Fundamentals. Indianapolis IN: Cisco Press, 2007.
[6] Juniper Networks, Network scaling with BGP labeled unicast, 2010.
[7] Kang Xi, H. Jonathan Chao, IP Fast Rerouting for Single Link/Node
Failure Recovery 2006.
Fig. 14 Bar chart for Test Cases.
41 [Link]