Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

OPTIMAL AND GUARANTEED ALTERNATIVE

PATH FOR MULTIPLE LINK FAILURES IN MPLS


NETWORKS
Siva Rama Krishna .A Mr.Ashutosh Kumar Dikshit, Asst. Prof
Dept. of Electronics and Communication Engineering Dept. of Electronics and Communication Engineering
VFSTR University VFSTR University
Vadlamudi, Guntur District, India Vadlamudi, Guntur District, India
sivaramakrishnafriend@gmail.com ashutoshcomes@gmail.com

I. Abstract--Traffic Engineering is one of the essentials


III. A.Label Information Base(LIB): It is the table
for backbones of Internet Service Providers(ISP).MPLS
contained in an LSR which gives the information of the
provides end to end performance for traffic by using
destination of a particular packet by examining given label.It
Label Switched Paths(LSP).Due to frequent failures in
is same as routing table in IP routing.
the network, the restoration of data has become a
challenging task to network administrators. We present
B.Label Switched Router (LSR):
basic recovery mechanisms, Makam's Scheme and
Haskin's Scheme compare them.The packet loss existed
in these mechanisms is overcome by introducing Fast It is the router which forwards packets based on the label
Rerouting and Reliable Fast Rerouting whose and by verifying Label Information Base(LIB).The
performance is far better compared to other two operation involved in the transfer of packets by LSR is
schemes, but packet delay and disordering exist.To swap operation.It is also called transit router.
overcome these, we propose Optimal and Guaranteed
Alternative Path recovery mechanism for multiple link c. Label Edged Router (LER):
failures. This is also called Ingress node or Egress node, which lies at
the end of the MPLS network.Ingress node will be at the
Indexterms--MPLS,Traffic beginning of MPLS network and Egress node will be at the
Engineering,LSP,Recovery,Ingress node,Egress Node end of the MPLS network.The label operation involved at
Failure,Backup LSP,Alternative LSP,RFR,OGAP (key the Ingress node is push operation, whereas the operation
words) involved at the Egress node is pop operation.

D.Label Switched Path (LSP): Label-switched paths


II. MPLS Introduction: (LSPs) is the path a packet chooses to travel along the
nodes.Each LSP is assigned a label value.There can be
Now a days it is common for backbone operators to use multiple LSPs in an
Multiprotocol Label Switching (MPLS) to route traffic MPLS network.
between Source and Destination in their networks. An
integrated approach to traffic engineering is provided by E. MPLS Header
MPLS traffic engineering for the backbone networks of
Internet Service providers.Multiprotocol Label
Label TC(QoS and ECN) Bottom of Stack TTL
Switching (MPLS) is one of the data-forwarding methods
for eminent realization of telecommunications This is also called Shim Header.MPLS functions by attaching
networks which forwards data from one network host to the an MPLS header to the packet, which contains one or more
next host on the basis of shortest path labels rather than labels(label stack).TC-Traffic Class, ECN-Explicit Congestion
long network addresses, to avoid complicated lookups in Notification, TTL-Time To Live.
a routing table(IP routing). MPLS works or lies in the
middle of conventional data link layer and network layer,
and thus belong to layer 2.5.
be higher if the ingress node is far distant from the point of
failure and frequency of packets transmitting is high.

B. HASKIN’S MODEL:
In this model, a backward LSP is set up from the point of
failure to the ingress LSR.Whenever a fault is identified by an
LSR which is upstream to the point of failure (POR), it sends
Fig: MPLS Network back the traffic along the same path(backward LSP) to ingress
LSR.Once the traffic reaches the ingress LSR, it is redirected
IV. MPLS Recovery: along the global recovery path.Here both the backward LSP
The recovery mechanisms can be classified based on two and the recovery path are pre-setup (protection switching).The
criteria, one is the recovery path followed and other is the time first packet reversed by Point of Repair (POR) along the
at which recovery is established. backward LSP is used as FIS signal.When the first packet is
The LSP from ingress LSR to egress LSR which forms the received by ingress LSR from backward LSP, the data packets
working path in an MPLS recovery network is called Working for primary LSP are rerouted into restoration LSP established
LSP or Primary LSP or Protection LSP. before from ingress LSR to egress LSR using a global recovery
The LSP which is rerouted to the alternative path after a failure mechanism.Until the FIS signal is reached to Ingress LSR, here
occurs is called alternative LSP or recovery LSP or restoration the packets originating at Ingress LSR experience a two-way
LSP. delay(Ingress LSR to POR and POR to Ingress LSR) before
Local Recovery: In this method, the recovery operation is being rerouted along the alternative path.The recovery process
performed locally to the link where failure is detected.The seizes when the last packet from sent by Ingress LSR is
intention of this type of repair is the fast recovery. received by it.The disadvantage with Haskin’s Scheme is the
Global Recovery: In this method, the recovery can be done delay in establishing recovery path and disordering of
if a failure occurs along any link of the protected path except packets.The disordering of packets is caused when the newly
Ingress and Egress nodes.In this type, all links and nodes use a originating packets from Ingress LSR combine with those
single recovery path or protection. arrive along backward LSP.
Rerouting: In this method, the traffic is redirected to the
alternative path after the failure is detected.Traffic is rerouted
on demand.The recovery path is based on fault information and
routing policies.The advantage with this type is no pre-reserved
resources are required and the disadvantage is it takes more
time than protection switching to establish recovery.
Protection Switching: In this method, the recovery path is
pre-calculated and pre-setup before the link failure occurs.Thus
immediately after the failure occurs the traffic is rerouted to the
pre-established recovery path.It is fast compared to rerouting
but it consumes resources like bandwidth.
Point of Repair(POR): It is LSR which detects the failure
and performs the recovery operation.
A. MAKAM’S MODEL:
In this mechanism, whenever link failure occurs the traffic
is rerouted using global restoration path, Once a fault is
identified in the primary path, a Fault Indication Signal (FIS) is C. FAST REROUTING MECHANISM
propagated along the LSP by POR to the ingress LSR, which is
responsible for rerouting.The ingress LSR should be intimated In fast rerouting mechanism whenever a POR detects a failure,
about the failure using FIS signal.The traffic along the LSP the packets are forwarded along the backward LSP.As soon as
before the FIS signal is sent to ingress LSR is lost.The loss will the LSRs in the backward LSP receives the packets from
POR, these LSRs begin to store the incoming packets from
ingress LSR in a local buffer which prevents the unwanted POR are sent back to the Ingress node using backward
transit of packets through the primary LSP(to and fro).Also, LSP.Any other packet coming from ingress node is also sent
the last packet forwarded by the LSRs is tagged to be back.This process avoids packet loss in link failure recovery.
identified when it is back in its way.This helps in avoiding Behavior of intermediate nodes in backward LSP:
disordering of packets when they are sent back by When a node on backward LSP detects a packet coming back
intermediate nodes to the ingress router. By tagging a packet, from POR, it takes it as FIS signal and sends it downstream
we can evade additional overhead.For this, we use ECN bits of the backward LSP to ingress router and invalidates the packets
the MPLS shim header. The packets stored in each LSR's stored in its buffer and the packets are sent to Ingress LSR for
buffer are sent back through backward LSP after it received routing into the alternative path.The further packet which is
tagged packet.The node which gave a tag to packet removes it coming next from the upstream LSR of primary LSP is given
when it returns.When the ingress node received its tagged a tag and is forwarded upto POR. All the following packets
packet the packets stored in its buffer are routed into that enter at this node along the primary path are stored in its
alternative LSP.The packets along backward LSP which buffer and not forwarded to next node.This reduces packet
reached ingress LSR are also routed to alternative LSP which delay by avoiding packets circulating in a loop of primary LSP
ends the restoration process.Then data packets which are in and backward LSP.
the buffer of ingress LSR and incoming data packets from IP Need of tagging in to avoid packet disorder
routers(source) are redirected to restoration path. By using When a node identifies a packet tagged coming along a
this mechanism we retain global reordering of packets, backward LSP, it learns that all downstream packets are
minimize packet loss. drained and now it must send back packets in its buffer.
The packets stored in each LSR’s buffer are sent back through
We can notice that re-ordering of packets is retained globally backward LSP after it received its tagged packet.The node
in the entire process.Even though Fast reroute scheme is better which gave a tag to packet removes it when it returns.When
than Haskin's scheme in reducing average delay in path the ingress node received its tagged packet the packets stored
recovery and in eliminating packet disordering, some premium in its buffer are routed into alternative LSP.The packets along
networks which require very low path restoration time will backward LSP which reached ingress LSR are also routed to
still get affected if they use fast rerouting scheme.So we alternative LSP which ends the restoration process. The
propose another model below. packets stored during this time in the ingress LSR, along with
all new incoming packets (from the source) are now sent via
D. RELIABLE AND FAST REROUTING(RFR) the alternative LSP. By using this mechanism we retain global
The purpose of RFR mechanism is to give Quality of Service reordering of packets, minimize packet loss and achieve fast
for data packets forwarded along primary LSPs in core restoration.
networks of MPLS.
We again consider the same scenario in which the backward V. MPLS Protection Domain (MPD):
and alternative LSPs are established.When POR finds an LSR A protection domain is a sequence of nodes bounded by
failed, it returns the traffic through backward LSP. The packet ingress LSR and Egress LSR (Primary LSP) including the
received first by ingress LSR from POR is treated as FIS restoration Path (Recovery LSP). If a protection domain is
signal.In this scheme, every LSR of the working LSP contains partitioned into multiple protection domains, then each
a local buffer and the data packet being forwarded to next segment is called a Segment Protection Domain (SPD). In an
LSR is stored in the buffer.There are two possible ways to SPD if a failure occurs, it is solved within the SPD (Local
store these packets to the local buffer.One is the non-swapped Recovery). Thus entire MPLS Domain is segmented into a no.
mode, in which the packets are stored in a buffer before the of MPLS SPDs.The SPD is an abstract of the autonomous
swapping of packets from working LSP to system in IP routing. SPDs are established according to
backward/alternative LSP is done.The second is swapped administrative routing policies. The bandwidth of alternative
mode, in which first swapping is done and then they are stored path for protection domain is used by Low Priority
in the buffer.Even though both modes work well, the non- LSPs(which has less priority than protected LSP) so that they
swapping mode introduces processing delay and additional can be preempted by high priority alternative path if a link
overhead in the LSP.In the non-swapped mode, if a failure is failure occurs
identified in the working path, the POR takes out the label
from shim header of the packet in the local buffer and then Path protection of an LSP is establishing an alternative path
perform the label swap operation for the corresponding output for entire LSP (Global recovery) when a failure occurs.
label and also changes the output interface.In contrast, in
swapped mode, the LSR takes the label from MPLS header of
the packet in the local buffer and sends it directly to the output
interface as the swapping is already done.This mode gives
better performance and is used in RFR mechanism.
Behaviour of POR:
Whenever a POR detects a fault, it starts switchover procedure
of packets,i.e., between working path LSR and backward LSR
by label swapping operation.All the packets in the buffer of
OPTIMAL AND GUARANTEED ALTERNATIVE We can use similar mechanisms to SPD3 for a fault in the
PATH (OGAP): link LSR3 – LSR4. If we use link protection mechanism,
For Single Link Failure: LSR3-LSR9-LSR10-LSR4-LSR5 provides the restoration for
The figure shown below is for OGAP with single link failure. Segment Protection Domain3.If we use path protection
mechanism, LSR3-LSR9-LSR10-LSR5 provides the
restoration for Segment Protection Domain3.
In the case of SPD3, the mechanism of path protection
establishes a short length recovery path than the mechanism
for link protection.
The OGAP for single failure combines the mechanism for
path protection with the mechanism for segment protection,
together with local restoration mechanisms which use the pre-
established alternative LSP for protected LSPs of the complete
MPLS network.The segment protection for each Segment
Protection Path functions in association with pre-established
alternative LSP, if the pre-established alternative LSP for
In this approach, we combine segment protection and path complete MPLS network is already is formed.This is because
protection with local recovery activation.The MPLS domain is the recovery path for the complete MPLS network is
divided into three SPDs. Each link shown in SPD in the figure established by joining of all the three Segment Protection
is a combination of one or more LSRs and links(not Domain recovery paths. The initial point of coincidence for
shown).The representation of different LSPs in MPLS both the segment protection domain of each segment and the
domain is given below: path protection for the entire MPLS network becomes the
combining point for the traffic along the rerouted path by each
The ingress node is LSR 0 and egress node is LSR 5. SPD into the pre-established recovery LSP. This mechanism
The working LSP or Protected LSP or Primary LSP is the makes use of link or path protection inside the SPD in
LSR0-LSR1-LSR2-LSR3-LSR4-LSR5 (solid lines) forwarding the data to the egress LSR (LSR5) of the complete
The pre-established Recovery LSP is the LSR0-LSR6-LSR7- MPLS network rather than carrying it to the respective SPD
LSR8-LSR9-LSR10-LSR5 (dim lines) egress LSR.
We shall implement OGAP scheme to the example in Figure.
MPLS Protection Domain (MPD) is LSR0-LSR1-LSR2-
When the link LSR0 -LSR1 is failed, the data is redirected by
LSR3-LSR4-LSR5 and LSR0-LSR6-LSR7-LSR8-LSR9-
LSR0 making use of the recovery path of SPD1. The initial
LSR10-LSR5
point of coincidence of the recovery path of Segment Path
Domain 1 (LSR0-LSR6-LSR7-LSR1) and the pre-established
SPD 1 is LSR0-LSR1, LSR0-LSR6-LSR7-LSR1
recovery path for the complete MPLS network LSR0-LSR6-
SPD 2 is LSR1-LSR2-LSR3 and LSR1-LSR7-LSR8-LSR9- LSR7-LSR8-LSR9-LSR10-LSR5 is LSR0.From this point of
LSR3 coincidence, the packets redirected by Segment Path Domain1
SPD 3 is LSR3-LSR4-LSR5 and LSR3-LSR9-LSR10-LSR5 make use of pre-established recovery LSP. So, LSR0-LSR6-
Backward LSP for Segment Protection Domain1 is LSR1- LSR7-LSR8-LSR9-LSR10-LSR5 forms the alternative path
LSR0 (dash-dotted line) for the complete MPLS network. If the link between LSR1
Backward LSP for Segment Protection Domain2 is LSR3- and LSR2 fails in Segment Protection Domain2, applying the
LSR2-LSR1 (dotted lines) same approach, the point of coincidence of LSR1-LSR7-
LSR8-LSR9-LSR3 and LSR0-LSR6-LSR7-LSR8-LSR9-
Backward LSP for Segment Protection Domain3 is LSR5- LSR10-LSR5 is LSR7. So, LSR0-LSR1-LSR7-LSR8-LSR9-
LSR4-LSR3 (dashed lines)  LSR10-LSR5 forms the alternative path for the complete
MPLS network. Lastly, if the link fails between LSR3-LSR4,
We can apply any type of protection mechanism inside an the initial point of coincidence of the recovery paths LSR3-
SPD, independent of other SPDs. To describe the approach, LSR9-LSR-10-LSR5 of Segment Protection Domain3, and
consider a fault in the link LSR0-LSR1 inside SPD1. When LSR0-LSR6-LSR7-LSR8-LSR9-LSR10-LSR5 is LSR9. So,
we use link protection mechanism, the restoration is provided LSR0-LSR1-LSR2-LSR3-LSR9-LSR10-LSR5 forms the
by LSR0-LSR6-LSR7-LSR1.When we use path protection alternative path for entire MPLS network.(see Table)
mechanism, the restoration is provided by LSR0-LSR6-LSR7-
LSR1.
We can use similar mechanisms to SPD2 for a fault in the
link LSR1 - LSR2. If we use link protection mechanism,
LSR1-LSR7-LSR8-LSR2-LSR3 provides the restoration for
Segment Protection Domain2. If we use path protection
mechanism, LSR1-LSR7-LSR8-LSR9-LSR3 provides the
restoration for Segment Protection Domain2.
Besides delay, the restoration path length indicates the
Table 1: Comparative Analysis of length of recovery path measure of resources needed to protect an LSP.
for single failed link of MPLS protection domain (from
ingress LSR0 to egress LSR5) In table2, we present the length of recovery path length using
link protection, path protection and OGAP for the complete
Faulty Link Link Path OGAP MPLS network (end-to-end) for multiple faulty links on the
Protection Protection basis of the network scenario we considered. By observing it,
within SPD OGAP scheme for multiple links requires 6 links only for a
LSR0-LSR1 0-6-7-1-2-3- 0-6-7-8-9-10- 0-6-7-8-9-10- restoration LSP, which has better performance compared to
in SPD1 4-5 (8 links) 4-5 (8 links) 5 (7 links) the path and link recovery mechanisms individually. We have
LSR1-LSR2 0-1-7-8-2-3- 0-1-7-8-9-3- 0-1-7-8-9-10- the working LSP length is 5 links.We can notice here that the
in SPD2 4-5 (8 links) 4-5 (8 links) 5 (7 links) length of restoration path in OGAP is just one link extra
LSR3-LSR4 0-1-2-3-9-10- 0-1-2-3-9-10- 0-1-2-3-9-10- compared to working LSP.
in SPD3 4-5 (8 links) 5 (7 links) 5 (7 links) We measure the percentage of packet loss and packet disorder
In the network, the actual length of working path end-to-end is using OMNeT++ tool.We simulate MPLS protection domain
5 links (LSR0-LSR1-LSR2-LSR3-LSR4-LSR5).In the above using Tknev tool of OMNeT++.The true performance of path
table, we compared the restoration path length between the restoration schemes is provided by the simulation by using a
ingress node and the egress node for a single faulty link. By file to plot the variables like packet loss, re-ordering, and
using the OGAP scheme for single failure we can provide a repetition. The file containing parameters to a plot is created
short alternative path length in comparison with other by programming in C++.The results are shown below:
schemes.It is better to implement the combination of segment
protection and path protection than to implement segment or
path protection individually.
For Multiple Link Failures:

The OGAP for multiple failures works similar to that of the


single failure.The OGAP for single failure can be extended to
propose OGAP for multiple failures.On the basis of segment
protection method, if the complete protected path is protected
from a faulty link in each Segment Protection Domain (i.e.,
more than one faulty link in working path), the length of
restoration path will increase because of repetition of the path
or link protection in SPDs.We observe the increase of
restoration path length if link recovery scheme is used.But, the
length of the restoration path will not be increased always if
path recovery scheme is used. The recovery path length of the
protection path is a major quantitative measurement of the
standard of a protection scheme. The length of recovery path
indicates the delay experienced by the redirected packets upon
a link failure. The above graph gives the results for multiple failures(2-3 and
4-5) when applied both of path protection with segment
Table2: Comparative Analysis of length of restoration protection. We can observe the packet loss of OGAP is zero,
path of multiple failed links for MPLS protection domain but 20% for Haskin Scheme and 55% for Makam's
Scheme.The re-ordering of packets in Haskin's Scheme is
Faulty links Link Path OGAP reduced to zero in the case of proposed OGAP.
Protection Protection
LSR1-LSR2 0-1-7-8-2-3- 0-1-7-8-9-3- 0-1-7-8-9-
and LSR3- 9-10-4-5 (10 9-10-3 (9 10-5 (7 links) VI . CONCLUSION:
LSR4 in links) links) . A long restoration time is the main problem of a dynamic
SPD2 and restoration scheme.To overcome this problem OGAP uses
SPD3 updated network information at the time of recovery to the
LSR0- 0-6-7-1-7-8- 0-6-7-1-7-8- 0-6-7-8-9- pre-established alternate LSP.This is possible by combining
LSR1,LSR1- 2-3-9-10-4-5 9-3-9-10-5 10-5 (7 links) the advantages of Fast Rerouting and Reliable and Fast
LSR2 and (12 links) (11 links) Rerouting.In addition to this, OGAP gives a guarantee of
LSR 3-LSR4
in recovery LSP at all times for the working LSP. OGAP also
SPD1,SPD2 provides protection of recovery LSP after the packets are
and SPD3 rerouted.OGAP considers multiple failures on an LSP and
gives shortest possible restoration path length and also the [9]. Application of MPLS-TP for transporting power system
failure of the new protected LSP.It also considers the protection data Steven M. Blair; Campbell D.
reversion operation.Reversion consists of redirecting the Booth; Jurgen Michielsen; Nilesh Joshi 2016 IEEE
packets from the recovery LSP to the primary protected LSP International Conference on Smart Grid Communications
after the failure becomes operational again.OGAP can be (SmartGridComm)
deployed for recovery in high traffic MPLS networks to [10]. Secure Inter-Cloud architecture for virtual cloud
achieve optimal performance. computing based on hybrid IP and MPLS infrastructure
solution Adao Boava; Yuzo Iano IEEE Latin America
Transactions 2016
REFERENCES.
[11]. MPLS multi-VRF design and implementation using GNS
. simulator Snehal Yadav; Amutha Jeyakumar 2016 IEEE
International Conference on Engineering and Technology
[1] MPLS-based hybridization in SDN Yash Sinha; Siddharth (ICETECH)
Bhatia; Virendra S Shekhawat 2017 Fourth International [12].Design of traffic engineered MPLS VPN for protected
Conference on Software Defined Systems (SDS)Year: 2017 traffic using GNS simulator Snehal Yadav; Amutha
[2] Leveraging MPLS Backup Paths for Distributed Energy- Jeyakumar 2016 International Conference on Wireless
Aware Traffic Engineering Frederic Francois; Ning Communications, Signal Processing and Networking
Wang; Klaus Moessner; Stylianos Georgoulas; IEEE (WiSPNET)
Transactions on Network and Service Management [13]. How to represent IPv6 forwarding tables on IPv4
[3] An analysis of multimedia traffic in the MPLS network in or MPLS data planes Sergey I. Nikolenko; Kirill
ns2 simulator Sergey Bukashkin; Marina Baranova; Aleksandr Kogan; Gábor Rétvári; Erika R. Bérczi-Kovács; Alexander
Saprykin Third International Scientific-Practical Conference Shalimov 2016 IEEE Conference on Computer
Problems of Infocommunications Science and Technology Communications Workshops (INFOCOM WKSHPS)
(PIC S&T) Year: 2016 [14]. Creating and managing dynamic MPLS tunnel by using
[4] Performance analysis of MPLS-based networks with SDN notion Hasanein Hasan; John Cosmas; Pavlos
conventional networks Junaid Tahir; Muhammad Zain Lazaridis; Sinan Khwandah 2016 International Conference
Siddiqi; Sharif Arif 2nd Workshop on Recent Trends in on Telecommunications and Multimedia (TEMU)
Telecommunications Research (RTTR)Year: 2017
[15].Real time distributed analysis of MPLS network logs for
[5].Upgrading Internet service provider (ISP) network in anomaly detection Muhammet Macit; Emrullah
multiprotocol label switching (MPLS) and border gateway Delibaş; Bahtiyar Karanlık; Alperen İnal; Tevfik Aytekin
protocol (BGP) environment Zhenxing Song; P. W. C.
NOMS 2016 - 2016 IEEE/IFIP Network Operations and
Prasad; Abeer Alsadoon; L. Pham; A. Elchouemi 2016
Management Symposium
International Conference on Advances in Electrical,
Electronic, and Systems Engineering (ICAEES)
[6] Internet Path Stability: Exploring the Impact
of MPLS Deployment Zakaria Al-Qudah; Mohammad
Alsarayreh; Ibrahim Jomhawy; Michael Rabinovich 2016
IEEE Global Communications Conference (GLOBECOM)
[7] Analysis of flows reduction scheme by adopting
two MPLS tags in software-defined network
Nattapong Kitsuwan; Eiji Oki 2016 IEEE Conference on
Standards for Communications and Networking (CSCN)
encoding algorithm for MPLS Segment Routing
Rabah Gue
[8] Label drez; Olivier Dugeon; Samer Lahoud; Géraldine
Texier 2016 IEEE 15th International Symposium on Network
Computing and Applications (NCA)

You might also like