A Diffserv-Aware Multi-Protocol Label Switching ...

CiiT International Journal of Networking and Communication Engineering, Vol 6, No 07, August 2014

279

A Diffserv-Aware Multi-Protocol Label Switching Traffic Engineering Applied on Virtual Private Networks Ayman E. A. Abdelaal, Fathi E. Abd El-Samie and Moawad I. Dessouky

Abstract---The demand on high bandwidth internet connections and applications is growing very fast due to the variety of internet arising usage either on business or entertainment. The Internet Service Providers (ISPs) have to enhance their internal networks in terms of bandwidth utilization and resources management. Two different technologies have been provided to help in solving the bandwidth utilization issues inside the ISPs networks, the diffserv Quality of Service (diffserv - QoS) and the Multi-Protocol Label Switching – Traffic Engineering (MPLS-TE). In this paper, the two technologies are used in conjunction with each other to get the highest performance, the best resources management and the efficient bandwidth utilization. Keywords---MPLS-TE; Diffserv-QoS; VPN. I.

INTRODUCTION

I

SPs have to deploy whatever technologies to ensure that they provide their customers with good services. One of the main problems face the ISPs is the bandwidth inside their network. If there is multiple paths inside the provider network to a certain destination, the routing protocol is responsible to choose the best path of them. There are many routing protocols like (Open Shortest Path First – OSPF, Intermediate System to Intermediate System – IS-IS, Routing Information Protocol – RIP, Border Gateway Protocol – BGP) and each protocol has each own metric to choose the best path. [3] [15] Some of them are depend on hop count like RIP, others has a configurable cost value like IS-IS, others has a complicated attributes like BGP [9] [10] and others are depend on the bandwidth on their metric calculation like OSPF. Even protocols which are using bandwidth as their metric still has a series problem which is that they use the overall bandwidth of the interface not the actual utilized bandwidth on it. That may lead that the protocol may choose a certain path (with bandwidth 100 Mega bit/second for example) as the best and route all the traffic and let another link with bandwidth 90 Mega bit/second idle because it is not the best which causes a congestion on the first link and under-utilization on the second link. MPLS-TE can help in solving the previous mentioned problem by override the default behavior of routing protocols.

Manuscript received on August 16, 2014, review completed on August 30, 2014 and revised on August 30, 2014. Ayman E. A. Abdelaal, E-Mail: [email protected] Fathi E. Abd El-Samie, E-Mail: [email protected] Moawad I. Dessouky, E-Mail: [email protected] Digital Object Identifier: NCE082014004.

0974-9713/CIIT–IJ-5389/07/$20/$100 © 2014 CiiT

It can choose the best path for a certain destination (sometimes called: Forwarding Equivalent Class - FEC) based on the available (un-reserved) bandwidth on the network links and the required bandwidth for that FEC. This model can influence the process of selecting the best path to a certain destination based on the reserved and un-reserved bandwidth but can’t limit or control the real traffic coming from the users which may lead also to a congestion if we defined the FEC to be 10 Mbps for example but the users uses it push a traffic of 50Mbps which lead to data queuing or dropping. As MPLS-TE is useful in reserving the expected bandwidth of each FEC but can’t control the real traffic of users which may exceed the expectations, the role of Diffserv QoS arises here. With Diffserv QoS we may avoid the congestion or manage the congestion or in other words control the real traffic of users by a process consists of several steps like: classification, marking, policing, shaping and scheduling. The rest of this paper is organized as follows: section 2 provides a short review on traditional network model. Section 3 provides a short review on the MPLS networks. Section 4 provides a short review on the Diffserv-QoS techniques. Section 5 presents the proposed model and the experiments results. Finally, conclusion is presented in section 6. II.

TRADITIONAL NETWORK

International Organization for Standardization (ISO) produced the Open Systems Interconnection (OSI) model to be the data network standard model and it is a 7 layered model. [1] [12] [13] Figure (1) illustrates the 7 layers and the common numbering of them.

Fig (1) OSI Model

Published by the Coimbatore Institute of Information Technology

CiiT International Journal of Networking and Communication Engineering, Vol 6, No 07, August 2014 Each layer of the 7 layers [2] has a function as follows: Physical Layer: Describe the physical signal like electrical and optical characteristics. Data Link Layer: Describe physical addressing (MAC Address – in Ethernet) which is link local, error detection and others. Many protocols work at this layer like (Ethernet, ATM, Frame-relay and others) Network Layer: Describe logical addressing which is global, internetwork communication, choosing best path to a certain destination and the most common protocol doing this task is “Internet Protocol(IP)” with its two versions 4 and 6. Transport Layer: Describe how to provide reliable transmission using segmentation, sequencing, acknowledgment, windowing, flow control and others. The upper three layers are special in application setting like managing sessions, code formatting, encryption, compression, and others. Encapsulation is the process in which the transmitter adds a header for each layer to perform the tasks associated with that layer. The de-encapsulation is the opposite process when the receiver removes a header of a certain layer after read the information on it and takes the appropriate needed actions. Each network element is called to be working on a certain based on which layer’s information it makes its forwarding decision. For example the Ethernet switch is called to be a layer 2 device because it forwards data based on the MAC address which is a layer 2 information and the router is called to be layer 3 device because it routes the data based on the layer 3 addresses. III.

MPLS technology could be summarized as [16] [17]: – Provides an intermediate encapsulation between an OSI Layer 3 IP header and an arbitrary OSI Layer 2 header – Can also encapsulate a non-IP payload – Bases forwarding on a label, regardless of the payload – Result: • Different protocols can be used to determine the path. • Different payloads can be used to provide different services. • Any traditional telco service or a functional equivalent can be implemented in an MPLS-enabled environment. MPLS Benefits [7] – Decreases forwarding overhead on core routers – Can support forwarding of non-IP protocols – Enhances BGP routing [6] ( no need for the internal provider routers to carry the whole internet routing table). – Supports multiple applications: • Unicast and multicast IP routing • VPN • Traffic engineering (TE) • QoS • AToM

IV.

MPLS (MULTI PROTOCOL LABEL SWITCHING)

[11] MPLS is introduced to solve many problems in the traditional network model and to add more features to the networking technology. The idea of the MPLS is to transfer the function of End-to-End transmission from the L3 to a new layer between layer 3 and layer 2 which is sometimes referred as L2.5.

QOS (QUALITY OF SERVICE)

Converged networks drive the need for QoS [18]when multiple applications supported over a common network infrastructure. Traffic from specific applications must be recognized and treated accordingly. Special handling is necessary to ensure that unique applications perform as expected in the face of congestion or queuing delays. User bandwidth usage must be controlled. Voice, Video and other critical applications cannot tolerate jitter and packet loss when they are in competition with data and other normal traffic.We need to manage the “unfairness”. Figure (3) shows a simple converged network.

Fig (2) MPLS Labeling

MPLS technology enhances IP routing in service provider core networks. Switching mechanism where packets are switched is based on labels. Labels usually correspond to destination IP networks. Only the routers on the edge of the MPLS domain perform routing lookup. An additional header, called the MPLS label, is inserted and used for MPLS switching. Figure (2) illustrates the process of label insertion.

0974-9713/CIIT–IJ-5389/07/$20/$100 © 2014 CiiT

280

Fig (3) Converged Network



281

QoS Parameters

–

WFQ (Weighted Fair Queuing)

Many parameters are used to measure the QoS of a network and the most important of them are listed below:[8] – Bandwidth: End-to-end information carrying capacity

–

CBWFQ Queuing)

–

LLQ (Low-Latency Queuing)

–

Delay: End-to-end delay for information delivery

–

Delay variation (jitter) – Variation in end-toend delays caused by packet queuing

–

Loss: Percentage of packets not delivered, usually related to congestion

The previous parameters could be improved by many ways. One of them is to improve the links bandwidth capacity but this is costly. Compression in data and headers could be used also but this application dependent capability. The modern adopted techniques are which use an advanced queuing strategy in the software queue of the interface and the below is a list of the most common of them: [8] – MDRR (Modified Deficit Round Robin)

A. Topology Explanation The topology is for a simple Service Provider includes from two Provider Edge (PE) routers which are (R1&R5) , three Provider (P) routers which are (R2, R3, R4), four Customer Edge (CE) switches which are SW1, SW2, SW3 and SW4 and each one of the CEs is attached either to data source or voice source. Layer 3 addressing protocol is Internet Protocol version 4 IPv4 or commonly mentioned as IP. IP address on the intraprovider links are from the private range 10.x.x.x and between the provider and the customer are used as 192.168.x.x .[4] [14] B. Tools i- Software GNS3 simulator v 0.8.6

0974-9713/CIIT–IJ-5389/07/$20/$100 © 2014 CiiT

(Class-Based

Weighted

Fair

Two definitions appeared to apply the QoS techniques: Integrated Services& Differentiated Services. V.

PROPOSED MODEL

From all the previous sections we can see that using only MPLS-TE on the network would be useful in resource utilization by reserve the required bandwidth for each data flow but can’t control the actual that may exceed the expected requirements. DiffServ QoS alone can control the amount of traffic on a certain interface, ensure some SLAs or prioritize some sort of traffic but couldn’t enhance the resource utilization of the network links which also may lead congestion problems. So the proposed Model is to use both of them in conjunction of each others. The below 5 experiments prove that the network performance with a simulated voice traffic is much better when use the proposed model. Figure (4) shows the topology used in those experiments.

Oracle VM VirtualBox 4.3.10 VMware Workstation 8 JunOS version 10.4 Windows XP professional service pack 3 JPerf SolarWinds Wan Killer ii- Hardware • LAPTOP Packard Bell: Intel® Core i5 CPU, 4 GB RAM. GNS3 is used because it is a software emulation enabling running the real Operating System of Juniper routers (JunOS) on the PC. Juniper is a leading vendor in the Data network field with powerful software & hardware products and is the main competitor of Cisco the famous vendor in that field.


CiiT International Journal of Networking and Communication Engineering, Vol 6, No 07, August 2014 Juniper is chosen because of the powerful of its systems. For the IP phones & PCs VMware is used as it is software enabling the installation of any virtual operating systems upper to the host operating system of the machine. Microsoft Windows XP is used as the guest operating System on the 4 virtual machines. Testing the characteristics of traffic is done by software called Jperf which can test the Bandwidth and Jitter and can emulate the voice traffic on the machines running as IP Phones. SolarWinds WAN Killer is used as source of traffic to push traffic from PC1 & PC2. Experiments target& Scenario Target is to get efficient bandwidth, delay and jitter between the two IP phones even when PC1 pushes traffic toward R4 utilizing link 2 and PC2 pushes traffic toward R5 utilizing links 2 & 3. Normal Routing(without TE) path selection choose the upper link as the best path which causes voice traffic to go through link 1,2 and 5 and link 2 is utilized with 2 sources of traffic from the 2 PCs and link 5 is utilized with traffic from PC2 only. Using Traffic Engineering will force voice to take the lower path as the upper one has no sufficient bandwidth. Still there are problems with link 5 as it is common between the two paths and here is the role of Diffserv QoS. [5] Experiment 1 Testing without any traffic from PC1 & PC2 Figures 5.a & 5.b illustrates two runs of the experiment and we can detect that the two IP phones could utilize 6 Mbps and Jitter was from 3-4 milliseconds. Experiment 2 In this experiment we push traffic from PC1 only without applying MPLS-TE or DiffServ QoS. Figures (6.a) & (6.b) are illustrated that when there is one source of traffic else of the measured voice source the bandwidth that the voice could get is about 1.5Mbps and the jitter reached 12.5 milliseconds. Experiment 3 In this experiment we push traffic from PC1 & PC2 without applying MPLS-TE or DiffServ QoS. Figures (7.a) & (7.b) show the results. When pushing traffic from both other sources we notice that bandwidth between the two IP phones is about 750Kbps and jitter is 25-30 milliseconds Experiment 4 In this experiment the MPLS-TE is applied when pushing traffic from the both sources but without applying the QoS. Figures (8.a) & (8.b) show the results. Some enhancement on the traffic pattern is observed when applying the MPLS-TE as bandwidth is about 1.5 Mbps and jitter is about 12 milliseconds.

0974-9713/CIIT–IJ-5389/07/$20/$100 © 2014 CiiT

282

Experiment 5 The last experiment which use the proposed model of the MPLS-TE and DiffServ QoS at the same time. Figures (9.a) & (9.b) show the results. The results of final experiment when using both MPLS-TE and DiffServ QoS shows end-to-end bandwidth of 5 Mbps and jitter about 7 milliseconds. These results are similar to the network status when it was idle without any congestion sources from PC1 or PC2. VI.

CONCLUSION

Best results could be gotten from using both MPLS-TE with DiffServ QoS in ISP and that guarantees providing customer with different with different Service of Level Agreement. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18]

Wendell Odom, “CCNA Routing and Switching ICND2 200-101 Official Cert Guide”, Cisco Press, May 2013. Wendell Odom, “CCENT/CCNA ICND1 100-101 Official Cert Guide”, Cisco Press April 2013 Wendell Odom, "CCNP Route 642-902 Official Certification Guide", Cisco Press, February 2010. David Hucaby, "CCNP SWITCH 642-813 Official Certification Guide" Cisco Press, February 2010. Kevin Wallace, “CCNP TSHOOT 642-832 Official Certification Guide“, Cisco Press, February 2010. Randy Zhang, Micah Bartell, “BGP Design and Implementation,” Cisco Press, December 2003. Luc De Ghein, “MPLS Fundamentals” Cisco PressNovember 2006. Cisco Systems Inc., “Implementing Cisco Quality ofService” Cisco Systems Inc., 2006. Cisco Systems Inc.,” Configuring BGP on Cisco Routers Volume 1”Cisco Systems Inc., 2005. Cisco Systems Inc., “Configuring BGP on Cisco Routers Volume 2,” Cisco Systems Inc. 2005. Cisco Systems Inc., “Implementing Cisco MPLS,” Cisco Systems Inc., 2003. Juniper Networks,”JNCIA-Junos Study Guide – Part 1” Juniper Networks. 2010. Juniper Networks,”JNCIA-Junos Study Guide – Part 2” Juniper Networks. 2010. Juniper Networks, “Junos Service Provider Switching“, Juniper Networks, 2010. Juniper Networks, “Junos Intermediate Routing,” Juniper Networks, 2010. Juniper Networks, “Junos MPLS and VPNs volume 1,” Juniper Networks, 2010. Juniper Networks, “Junos MPLS and VPNs volume 2,” Juniper Networks, 2010. Juniper Networks, “Junos Class of Service, “ Juniper Networks, 2010.



0974-9713/CIIT–IJ-5389/07/$20/$100 © 2014 CiiT


283


0974-9713/CIIT–IJ-5389/07/$20/$100 © 2014 CiiT


284


Ayman E. A. Abdelaal received the BSc degree from the faculty of Electronic Engineering from Menoufia University in Egypt in 2010. He works as a professional Engineer and a freelancer instructor in many academies delivering network courses. His major interest is the network filed including the routing & switching and service provider applications.

Moawad I. Dessouky received the BSc and MSc degrees in Electrical Engineering from Menoufia University in Egypt in 1976 and 1981 respectively. He received the PhD degree from McMaster University in 1987. He is currently the vice dean of the Faculty of Electronic Engineering, Menoufia University. His areas of interests are signal processing, Image Processing and Satellite Communications.

Fathi E. Abd El-Samie received the B.Sc. (Honors), M.Sc., and PhD. from the Faculty of Electronic Engineering, Menoufia University, Menouf, Egypt, in 1998, 2001, and 2005, respectively. He joined the teaching staff of the Department of Electronics and Electrical Communications, Faculty of Electronic Engineering, Menoufia University, Menouf, Egypt, in 2005. He is a co-author of about 130 papers in national and international conference proceedings and journals. He has received the most cited paper award from Digital Signal Processing journal for 2008. His current research areas of interest include image enhancement, image restoration, image interpolation, super resolution reconstruction of images, data hiding, multimedia communications, medical image processing, optical signal processing, and digital communications.

0974-9713/CIIT–IJ-5389/07/$20/$100 © 2014 CiiT


285