WAN Network Design - Assignment Example

Vin da France WAN Design Being seperated by distance is a challenge for a growing company. Expanding a remote area office is extremely needed to cut transportation fee, the employment matters, and other necessity. However, we have to establish communication connectivity between these remote areas and their headquarters to maintain the company's important data correspondences. Some technologies which can be used for WAN including frame-relay link, ATM circuits, or pure clear-channel network. These technologies combined with IP as the overlay network would be a good choice since IP network can deliver appropriate any-to-any network. Many telecommunication carriers offer leased circuit as a solution for this distance barrier. Determining the link bandwidth and topology of the network should be an important subject to discuss. To minimize operational fee by expanding a remote office is our main goal. Yet, we will unable to achieve this objective if we miscalculate the network development cost, although, we eventually have the ability to deliver data communication among the nodes (we often call a network entity as a 'node'). WlvNET software WlvNET is software that enables us to calculate network development cost for each topology that we simulate. By knowing the cost, we will able to find out the best network path that meet the maximum cost requirement for the company. What wlvNET do is basically an iteration of an algorithm to find the shortest-path of a link related to the incoming/outgoing traffic of a node. It uses Djikstra's algorithm as this method has been used in most of shortest-path calculation (lecture notes, sli. 13). Any topology possibilities could be shown by this software. Unfortunately, we have to decide the optimum network topology manually. This would be a very essential issue. One of the parameter that this tool uses in its iteration is incoming/outgoing traffic. This parameter need deep discussion as it will effect the link needed to overcome the traffic. One thing should be taken care of is that upload and download traffic have a ratio which depend on the application used by the end-point/user. For example, simple usage of internet to browsing, e-mail, and instant-messenging would give an upload/download traffic ratio 1:8. Other type of usage, such as LAN-emulation and banking application using frame-relay connectivity will yield a ratio of 1:3. Moreover, a voice application, such as voice over IP has a ratio of 1:1 for its upload/download traffic. These measures based on the traffic pattern tendency that my company record from the customer's private link. The second variable that important for this software to run appropriately is, of course, the distance between any nodes. Our main goal is to find the minimum cost for sum of the entire link. Hence, cost per kilometer for a link is a very crucial issue. Distance between the nodes could be achieved by using simple coordinate calculation. I use a plain method where I map the cities of Vin da France network to (X, Y) coordinates and subsequently, scaling of the distance will depend to wlvNET calculation mechanism. I have tried several experiments and I come to a conclusion that wlvNET will automatically scaling the distance based on the coordinates as the input. Hence, the cost of the network will remain the same although we increase the distance value for a link. The cost will be different if we make a correction on the coordinate position. Another optional variable is the personnel population of each node. This variable will result to final traffic calculation that going from/to each node. Actually, we need to determine the distribution traffic pattern for each personnel's traffic. However, due to limitation of each user's statistical data, I take an assumption that each personnel will have an equal daily traffic. One weakness of this application is that we could not simulate multiple line type to be used in the network. We only able to specify one line type for each single simulation, whereas combination of line types will result a more cost-efficient network. The reason is because multiple line type will handle the traffic fit exactly as bandwidth demand. So, there will be no useless portion of bandwidth. Methodology To provide sustainable network configuration, each link of the network should not over-utilized. Moreover, there should be a back-up link which able to handle the traffic in case a link is fails. Hence, we determine that the average utilization of the entire network link is 50 %. This is an ideal value to deliver a full redundant network. I take an assumption that the upload to download ratio for the network is 1:2. With this criterion as our directive, the wlvNET could process the simulation then. The table below is summary of the experiments using wlvNET. Each cell with blue-colored font is the network cost for particular line type and alpha. Line Type Alpha 128K 256K 512K 0.04 822,876.7 421,554.5 283,229 0.05 764,244.9 402,736.8 274,545.7 0.06 708,823.3 384,109 248,403.6 0.07 549,158.9 297,321.7 191,258.1 Alpha is a variable that indicates the deviation of the network central node/city. Increasing the alpha will yield a network that more centralized. A star topology would fit an example of centralized network. Alpha ranges from 0 to 1 with 0.01 increments. From the experiment results, it shows that the simulation result a same output starts from alpha equal to 0.07 and above. The previous table shows that using link with higher bandwidth line type could decrease the cost significantly. The reason for this outcome is because higher bandwidth line type could replace several lower bandwidth line types. For example, four of 128K line type can be replaced with single 512K and return a link cost that less expensive. That is what happens when we use 512K line type instead of the 128K on the wlvNET simulation. Since there is no sufficient material and documentation about the cost calculation algorithm of this software, I take an assumption that the cost reported is the actual cost which has already been scaled. For example, a network simulation design of 256K line type and alpha number 0.5 would cost 402,736 euros. Recommended WAN Model From the table on previous section, I choose three network topologies which have minimum cost. All of them use 512K line type. The alpha number for these models are, respectively, 0,05; 0.06; and 0.07. I capture the utilization graphic instead of multiplicity line type graphic to show the traffic record of each link. Yellow lines indicate that the link is utilized under 30 %, red undotted line for traffic under 50 %, and red dotted line for less then 60 % utilization. I place the snapshot of the network topology alternatives on the appendix section. The first alternative (alpha = 0.05) offers a simple configuration. It provides redundancy for entire city, except Strasbourg. However, we can add single redundant link from Strasbourg to Dijon, for example, if redundancy issue is so necessary. The utilization is also adequate to be handled. There are no links that under or over-utilized. I would recommend this model as the most appropriate WAN design. All of the links are single 512K line type, except Paris - Bordeaux link which is a bundle of three 512Ks. The second alternative (alpha = 0.06) slightly different from the first one. This model is actually just a modification of the first network model. With several link eliminated and replaced with shorter link, the network cost decrease not so significant. There is a decreasing of 26 thousand euros from cost of the previous model. The third alternative (alpha = 0.07), simplify the network to more compact network. It offers the most minimum cost. There are two network 'hemisphere', the north and south part, with Lyon as the focus point. Although this model has the 'cheapest' development cost, it potentially unsafe for long-term operational condition. If, for any reason, Lyon failed or let us say the router or switching device on Lyon damaged or malfunction, the north and south hemisphere will be unable to communicate. Another issue for this centralized topology is that it will consume too many device interface/ports in Lyon. From the diagram, we could see that there are six links come into Lyon. Hence, the device should has 6 port connected to other nodes. Solution for this Lyon-centric network is that we have to provide system or device redundancy instead of link redundancy in Lyon office. By concentrating on our system, we could trust the link matters to the telco carrier. In fact, an availibility of a link is part of carrier responsibility. The customers that lease the link could demand their right based on the service level agreement with the carrier. Conclusion I would prefer to choose the first WAN design. Beside it has low development cost, it gives high network reliability with its link redundancy. It is necessary to keep thought in mind that we better invest on higher reliable network which also more expensive cost, for a link failure will cause great matter of financial loss. Works Cited Index. Wolverhampton Network Design and Security class's website. University of Wolverhampton. 29 October 2007. Appendix Read More