This paper engag es the utilization of MPLS in an Internet service provider (ISP) environment. Furthermore, the knowledge of what MPLS stands for, how it is used, and which characteristics it has, will be expanded. Therefore, a SWOT analysis will reveal its strengths, we aknesses, opportunities and threats. Keywords ”MPLS ; ISP; G.8032; VLAN I. INTRODUCTION MPLS, short form of Multi -Protocol Label Switch, is a packet transport technology. It works between the second and third OSI layer [1], whereat the second one deals with the frames, and the third one with the IP.
This is the case, because it cannot opera te in just one of these two layers, therefore MPLS is often declared to work between those two layers “in the layer 2,5. Just like the technologies IEEE 802.1Q and IEEE 802.1ad, which describe the utilization of VLANs and Q in Q, MPLS also creates a head er between the ethernet and an IP -header. In literature, these headers are described as MPLS labels.
[2] Both VLAN and MPLS labels have the ability to stack more than just one packet. With these following paragraphs the strengths of the Multi – Protoco l Label Switch are going to be ANALYZED and struc- tured, starting with the most important one. II. STRENGTHS If a closer look is taken at the layer three, the routing, it shows that every single packet has to be analyzed before transferring it. It is necessary to determine the best possible route for every packet. In order to do so, it has to have an overview of the whole n etwork. Routing can be abstracted as a hop -by -hop based protocol. Each packet needs to be analyzed be each of the intermediated devices. A downside effect would be that more resource power is used even though, it would not be needed. As an ex-ample, imagi ne that two packets, with the same transmitter ad- dress as well with the same delivering address would appear right after one another. The routing layers would analy ze for each packet the best possible route. [3] This recalculation can be considered as a wasted effort and therefore as an ineffi- ciency. Furthermore, to generally place a router in a network is by far more expensive than his competitor, the MPLS – Switch. This downside effect has been evaluated and considered while designing MPLS. Inste ad that every packet gets its suitable route evaluated, one connection will get labelled. Based on this label, it will be determined where to forward the packet. If a MPLS receives a packet it will consider its label -table to find out where to pass it on. During the reception and the transfer of a packet to the next hop, the label will be removed and re-placed with a new suitable one before sending it on. [4] There-fore, the label on a MPLS -switch is unique on each device. However, this does not apply for VL ANs. One connection in an ISP represents one costumer, and therefore this VLAN must differ from all the others in the whole network. TABLE I. MPL S Label MPLS Label 20bit Exp 3bit S 1bit TTL 8bit VLAN Tag TPID 16bit User Priority 3bit CFI 1bit VLAN ID 12bit a. Comparison between VLAN Tag and MPLS Label In Table1 we can see a comparison between the VLAN -Tag and the MPLS -Label. It is striking to see that both have the same length. In both cases the packet will be expanded by 4 – byte, which are 32 bits. Furthermore, MPLS enables a bigger addressing scheme t han VLAN, where for addressing only 12 bits are used compared to MPLS, whereat 20bits are necessary. [2] The re are also several issues if an ISP runs its network without a MPLS : huge MAC t ables on Switches double -tag packet s Assuming a customer would have a very flat network this would indicate that they have access to huge broadcasting do-main to which they can connect dir ectly. This case would arise when an ISP connects two campus net-works with each other. Every switch of the ISP would have a big Forward Data Base (FDB), where the to be reached MAC addresses with their interfaces are stored. Therefore, an enormous load o f MAC addresses would always point to the same interface. This described issue would only apply for just one customer, however, an ISP has more than just one, this MPLS for ISPs Andrey Cvitkovic Telecommunications and Internet Technologies FH Technikum Wien Wien, Austria [email protected] -wien.atwould result in a much higher need for resources in storage and processing power. Not all Switches support a very big FDB and at some point, the ISP would have to restrict the MAC learning process for the customer. This countermeasure would, on other hand, come with negative effects for the customer. Nevertheless, it is the only option, because this would also affect the core of the ISP. Every packet needs to be analysed in order to determine at which interface it needs to be transmitte d to. A following c onsequence would be that passing on packets would take longer than initially calculate d and therefore de- lays might appear. This is the case, because the FDB is a list and to process it from top to bottom the time may vary – de- pending on how many things are listed, it may take some times longer and sometimes shorter. A result of this would b e Jitter . Jitter describes the behav- iour when packet s don’t arrive at the same speed. Sometimes they arrive sooner and sometimes later than expected. Delays and Jitter are destructi ve for Voice over IP (V oIP) . [5] The ISP would not be able to deliver a good service to the cus- tomer. A telephone cal l would get disconnects and the voice quality would be hugely disturbed. Depending on the sold Ser- vice -Level -Agr eement (SLA) this could mean a violation to the agreement and be a s erious matter for the ISP which could end in penalties. To better fu ll the SLA an international committee , named Metro -Ethernet -Forum (MEF ) [6] , has been formed to cer- tificate har dware and to standardize protocol s. This co mmittee took the comple x MPLS and designed new subcategories to consider the special needs of IS P. Therefore MPLS -TP , and dynamic MPLS -TP ha ve been develo ped. MPLS has for each of the two communication partners a uni- directional tunnel with a set of label s across the network. MPLS TP , a subcategor y of MPLS , guaranties that the paths can be used concurrently and therefore mo re eff iciently. In- stead of using diffe rent communi cation tunnels, like MPLS, the MPLS -TP uses only congruent paths. [7] This would further guaranty a symmetrical behaviour in both directions. This is not only important for the ISP site of perspective but also for the customer ‘s. This comes in handy for fault management and the visibility in the network itself. Another for m of MPLS which comes in use by ISP’s , is dynamic MPLS -TP. The difference here is that the co nnection can react to a topological change. [7] With this method uptime of a connection can be drastically improved becau se if a connection in the network fails , a new route will be taken in order to ensure a save transfer of the packets. These incidents could occur if a Switch stops working o r maybe a fibre connection of one network element to the other has been cut during construction work. III. WEAKNESSES One of the biggest disadvantages of MPLS is the complexity which appears in troubleshooting and maintaining a MPLS networks. [7] It is undeniable that MPLS is far more complex and s upports more features than its competitors. On one hand, customers are always craving for new features which in the end should improve the offering of the ISP with stability or ease of usage. But these kinds of features require more time, especially for troubleshooting, because the problem may also be more complex. In these kinds of new features often bugs and malfunctions are hidden. Nevertheless, the customers pay more for their special features that’s why it is not easy to neglect new features. There is also a downside for ISP, which does not normally start with MPLS but evolves into it. The reason for that are economical aspects. Hardware nowadays support mostly MPLS and its major competitors. Unfortunately, when an ISP wants to activate MPLS netwo rk it has to buy the licenses for it. These are expensive, especially when you take in account that for every network element a license has to be bought. When an ISP is economically big enough to move to MPLS there is always the downside that the whole em ployees have to be retrained. Mostly an approach is taken, where a parallel operation with MPLS and its presuccessor is used. [8] With this a successfully migration can be guaranteed. IV. OPPORTUNITIES Big opportunities for MPLS generally lye in big transport networks r egardless if they partially act as a small ISP or not. An ideal utilization of it would be in the sector of energy suppliers. The have enormous networks which they can futureproof with MPLS. Energy suppliers mostly have old technology in use for their pl ants. One of these would be Plesiochrone Digitale Hierarchie (PDH) or Synchrone Digitale Hierarchie (SDH). These are described in [4] and [2]. They could without problems still use their connections but transport them with MPLS. With this approach, they have the possibility to change to another technology without having to change their entire infrastructure at any point in the future. This would also economically make sense, because the energy suppliers do not have to make another big investment in order to change to a different technology. They have the chance to slowly pl an and take every doubt into consideration and with a good performing project management team they have the opportunity to invest and install newer technologies. This could be derived because old technology cost more in maintaining. Only support -costs woul d increase drastically yearly and they would have less and less selection for the support provider each year. On one side, would the supporter keep the loyalty to their customers, because they can ask more for support each year, because the employees would have to deal with it and also could not participate.Another problem would be the organization for spare parts if they get broken. These would be hard to get and also potentially be more expensive, which is a problem for energy suppliers. They have to have a very stable long -lasting connection with backup li nes, in case of a replacement, when the primary one fails. V. THREATS The Biggest threat to MPLS for ISP is G.8032, also known as Ethernet Ring Protection. Especial ly for smaller ISP this option appeals to be very attractive, because it is by far cheaper to realize and the license cost are restricted as well. It is easier to maintain and troubleshoot, because it has a simple topology, which is easier to understand an d act accordingly if any problems occur. In controversy to MPLS, G.8032 is not based on a mesh architecture but on a ring one, as the already name indicates. Such a network has in its core a major ring on which subrings are attached to. A separation of co stumer traffic is done with VLANS instead of a MPLS label. IEEE 802.1ad, also known as Q in Q, is therefore used. The rings are physically closed but virtually one link is always deactivated or on standby, because it activates itself, if one of the links i n the ring fails. One disadvantage of that is that each ring can only absorb one failure. [9] There is always an unused connection in the network which only accounts as a backup and blocks a connection before its failure. Luckily a workaround already exis ts. It is possible to create more than one logical ring on a physical one. The blocking links only occur on the logical ring. It also is possible to set more logical rin gs with different blocking links. With this, customers could be handled and interconnected, which would take in account that an even coverage of the whole ring is possible. CONCLUSION MPLS has its advantages like a more efficient and faster way to transmitter packets Therefore , less delays a nd a better service availability can be guaranteed . However, companies need to consider t he license cost for the MPLS. It wi ll be interes ting to see how MPLS can be improved and which transport protocol will be establish ed in an ISP environment in the future . REFERENCES [1] O. Obafemi, T. Gyires und Y. Tang, ћAn Analytic and Experimental Study on the Impact of Jitter Playout Buffer on the, in The Tenth International Conference on Networks , Illinois, 2011. [2] E. Uchytil, “Carrier Ethernet and optical networks,” 2017. [3] W. Hickey, ћ Ciena, 12 June 2017. [Online]. Available: -MPLS -TP- in-the -Metro -Network -Why.html. [Zugriff am 4 1 2019]. [4] T. Robertazzi, ћMultiprotocol Label Switching (MPLS), in Guide to Reliable Internet Services and Applications , New York, Springer, 2012, pp. 53 -57. [5] T. R. Tronco, ћEvolution of Internet Architecture, in New Network Architectures: The Path to the Future Internet , Berlin Heidelberg, Springer, 2010, pp. 25 -55. [6] A. Dumka und P. H. Lal Mandoria, ћDynamic MPLS with Feedback, International Journal of Computer Science, Engineering and Applications (IJCSEA), pp. 125 -130, April 2012. [7] C. R. Kalm anek, S. Misra und Y. R. Yang, ћ”Structural Overview of ISP Networks” in Guide to Reliable Internet Services and Applications, in Guide to Reliable Internet Services and Applications , London, Springer, 2010, pp. 19 -93. [8] M. Z. Stanikzai, ћ -stanikzai.com, [Online]. Available: -stanikzai.com/mpls -backbone -migration -steps/. [Zugriff am 7 1 2019]. [9] Ciena Corporation, ћCarrier Ethernet Associate (CE -A), 2010 -2015.