KINGSTON UNIVERSITY
Network Design and Management Coursework CIM238 Amarpreet Singh Saini K1051678 K1051678 6/17/2011
A company in Greece named Gamling Printing CO. asked out team to design a network based on specifications and requirements provided by them. Based o n the requirements of the client our team designed a network which is reliable, secure, manageable and robust.
1. Executive Summary This report is done on behalf of Gamling Printing Co. for the purposes of designing a network based on the requirements and specifications provided by the organisation. The information contained in this document is confidential and sensitive and should be handled with care. Based on the requirements specified by the organisation, we have designed a network which fulfils and serves all the requirements outlined and detailed, which is secure, managed, reliable, adaptable and cost effective. It not only meets the current demands but also provides the capacity for future growth. Furthermore standards and recommendations recommendations of feasible technology are also outlined in this document. The proposed network design offers a number of advantages over the currently utilized network design. -
Ensures that the network infrastructure is well positioned so as to avoid physical disasters.
-
Network infrastructure that is more reliable, flexible and cost-effective.
-
A network that can meet the current demands put forth by the organisation and has the capacity for future growth.
-
Provides consistent improvements in services.
-
Ensures a level of robustness required by the organisation for the efficient functioning of its business.
The proposed network can be implemented im plemented with minimal disruption and less downtime. Below is a figure illustrating the process of migrating to the new network design.
Fig. 1
The new network has the following key features in terms of technical solutions implemented. -
The terminals are wired into a 1Gbps LAN due to the high bandwidth requirement of the internal network.
-
The hardware is very cheap and affordable and a single server rack is all that is used to setup the network. This is very beneficial when backups are taken into consideration.
-
The design is maintenance free and requires very little technical expertise to o perate and even diagnose due to its simplistic nature. It uses a single Application/Domain controller and a Network Attached Storage that is critical to the functioning of the network.
2. Requirement Analysis Business Requirements The network upgrade as specified by Gamling Printers Co is required to: -
Update the outdated and aging infrastructure Minimize the cost of the network infrastructure Support future business expansion Increase productivity Offer flexibility and scalability in terms of bandwidth required for various kinds of services and differing business traffic types Add a level of security to the business-specific and sensitive information transmitted. Guarantee minimal delay for applications that are critical to the organisation A highly available network that can help minimizing planned and unplanned downtime. Be ready to embrace new technologies in the future as required by the business. Offer alternative and multiple means of access such as wireless LAN Disaster recovery and redundancy and appropriate mechanisms to mitigate consequences.
The organisation runs many services on the workstations. Due to the nature of the business as a printing facility they run photo editing and design software on all the machines. The services are to be discussed in detail in design section of this document but for the purposes of the requirement specification we list them here. These are HTTP for web data, Database Query and Fetching and FTP to access the Network Attached Storage (NAS), Print. It can be noted that the internal LAN needs a high speed channel whereas the organisation can opt to go for a less fast line for connecting to the internet as their HTTP usage is only medium load.
Existing network Infrastructure The various deficiencies and risks associated with the current network employed by the o rganisation are: -
-
The current network is highly rigid to maintain because it does not conform to standards. There are many different technologies used which are very difficult to manage. Connectivity options like as coaxial cable is a big constraint. Due to ado ption of such legacy options much of the time is wasted on resolving solutions that can be avoided in the first place by having a standard conforming model. The cost to maintain the current is excessive and drains a lot of capital than it should be. The existing network can’t handle the quick deployment of new services required by t he organisation. Absence of a strong security solution is a big threat to the organisation’s assets. There is no security policy implemented at the organisation premises. Lack of fault tolerance.
3. Design and Architecture To construct the new design for the network we used many techniques and approaches. We now discuss them briefly. Studying the existing network The existing network design was studied, analysed and documented. Various aspects of the network were studied and investigated. Some of those are- the number and speed of access points, transmission channel information, details of applications and services used, and environmental aspects as such. Discussion Discussions were held so that various constraints, business requirements and future directions could be taken into consideration. Requirements were documented and used as the building foundation for the design Traffic Analysis Data flow and volume were gathered for recognising various bottlenecks that could be affecting the existing setup and also prioritizing the areas that were more critical Migration Plan Development of a migration plan occurred as well by which the proposed design could be implemented smoothly so that the networks remains highly available during the transition.
There were some aspects of the network that were however excluded and not taken into consideration while designing the network. These included an interactive backup solution, which as told by the organisation management can be handled by them manually. Also there is only one link for connecting to the internet.
The proposed network design and Considerations -
The terminals will in a centrally switched infrastructure which will allow for easier management. This allows for better network performance as the link used within the internal network is a 1 Gbps line.
-
Both the servers and the terminals will have a better access speed in the network than the existing network setup. Redundancy is also available in case of an outage of failure of network components.
-
-
Since we use single vendor products namely Cisco routers, switches and HP servers and terminals, management is substantially easier. Wireless access is also possible. Right now the service is used for internal wireless printing services but it can be configured to provide internet access to customers or internal employees. High availability depends on both the services and the whole network. Although MTTF is a factor, availability usually depends on the network. Hence adding redundancy is a good investment and resources should be allocated for it.
Fig. 2 Network design
The network design consists of a HP server that is application and domain controller, a Cisco router, switch and wireless AP. The firewall again is a Cisco product with the terminals along with the printers and scanners being HP products again. The list of equipment is listed in the figure below.
Fig. 3 List and number of equipment employed
We have an ATM e3 line connected to an ADSL modem which is further connected to a router. The router is connected to a Cisco hardware firewall solution, which is further connected to a switch. This 1 Gbps switch is connected to the Domain controller/Application server, the Network Attached Storage, the wireless printers and the terminals. The terminals are connected to the switch using UTP Cat6 cable whereas the printers and plotters are wirelessly connected via the means of and wireless Access Point. We now outline some of our design choices. -
The 1 Gbps switch is only for the internal network as simultaneous requests are made to the network attached storage. Plus the huge file sizes make the internal network bandwidth intensive.
-
The printers are wireless as due to intensive use their fail often. In such a case employees can use the printers near to them due to their wireless nature. This saves a lot of time and effort.
-
The Network attached storage is a highly beneficial investment as it can be easily backed up manually as per the organisation’s requirements. A central storage offers a lot of benefit than individual storage devices on the terminals.
In the figure below we show the actual floor plan with the server rack used on the organisation’s premises.
Fig. 4 Floor plan along with server rack diagram
For the sake of demonstration we have only employed 4 sets of terminals and devices. In the actual scenario we have more terminals than those. The UTP Cat6 cable runs along the length of the premises. Each workstation has a wireless printer next 2 it. There are 2 wireless plotters on the floor.
4. Specification The availability of the network link is explained below. The mean time to repair as given by the organisation is 3 minutes. Using MTTR as 3 minutes we derive availability as follows.
Availability= MTBF/ (MTBF+MTTR) MTTR=3 minutes*52 weeks=106 Agreed service uptime = 24*60*365=525600 MTBF= (Agreed service uptime- MTTR) =525600-106=525494 Availability= MTBF/ (MTBF+MTTR) = (525494/525600) =0.9997983257 The overall availability percentage is 99%
We calculate the traffic generated over the internal network in the following manner. Traffic in megabytes per day = EmployeePerDay * RequestPerEmployee * PageSize / 1048576
The 10 employees each handle around 70 jobs every day generating a data of about 50000000 bytes on an average request. 33378.60107421875 Megabyte per day or 991.2531822919846 Gigabyte per month
5. Simulation
For the purposes of the simulation which we performed in OPNET we used the following network design.
There is an ATM e3 line connected to an ADSL Modem, to a router to a firewall. The firewall is connected to a 16 port 1Gbps switch. The switch is connected to a single server that serves the dual purpose of NAS and Domain controller. There are 10 workstation connected to the switch. We have excluded the wireless products out of the simulation.
The simulation ran for 1 day during which no traff ic was dropped as shown in the simulation result.
The above figure shows detailed statistics fo r the network health. As said before our model dropped no traffic.
The above figure shows the delay in seconds the Ethernet. As we can see that this is very low, only peaking less than .000035 which is a very good value.
The above simulation result shows the response time for the DB Query service along with traffic sent and received in packets per second and bytes per second.
The above screenshot shows the TCP delay in seconds for a whole day which again is very low.
The above figure shows the simulation results for TCP Segment delay.
Group Coursework Presentation Athanasios Politis Amarpreet S Saini Alkiviadis Tsitsigkos
Small size business
Copy Centre Business
5 employees
Plain Home Network Utilisation
Fast & Reliable Network
Secure Network Storage
Wireless connectivity
Improve Network Security to avoid data leakage on Internet
Wired Terminals, Wireless Printers
Cheap & affordable hardware
Maintenance FREE design.
Capable of expanding in near future
More secure & robust than it used to be
Had a Home Network Implementation and now has a Small Business Network Implementation
Business is more agile with Wireless connectivity now
Future Expansion Capability
Wireless Link Dimensioning 1. Abstract Due to the need of many quality-of-service requirements of voice and data in GSM/GPRS networks, channel allocation techniques are very vital as far as system performance gains are concerned. A study on the paper by Marques and Bonatti where they present and compare o f two techniques for wireless link dimensioning of telecommunications networks is discussed in this paper. The techniques are based on Markovian model and priority model. A brief introduction of GSM/GPRS is also presented along with a look at channel allocation schemes by other researchers. Key Terms – Wireless link dimensioning, Channel allocation, Markov, Priority, Sharing.
2. Introduction The integration of GPRS and GSM is a significant change that has hushed in a new era in the wireless networking and communication field. Due to GPRS users can benefit from faster data rates than before with much short wait times. GPRS enables easier access to packet switched networks in a wireless communication medium. Due to the vast and crucial need of many quality-of-service requirements of voice and data in a GSM/GPRS network, channel allocation techniques are highly important as one can achieve a significant increase in system performance. [1] Here in the figure below we can see an implementation of GPRS over a GSM network.
Fig. 1 GPRS over GSM [2]
In a GSM network, a mobile station (MS) communicates with a base station subsystem (BSS) via the wireless communication link. The base station is connected to the mobile switching centre ( MSC)
which further interacts communicates with the visitor location register (VLR) and home location register (HLR) to keep a track the locations of the mobile stations. [2][3] In the GPRS/GSM architecture, MS, BSS, VLR, and HLR are structured differently. For e.g., the home location register is further tuned to accommodate GPRS user data such as credentials. Two GPRS support nodes (GSNs), a serving GPRS support node (SGSN) and a gateway GPRS support node (GGSN) are present in GSM/GPRS architecture. The SGSN makes sure that the mobile stations receive the packets intended for them. The gateway GPRS support node acts as a gateway between the GPRS and external networks. GPRS tunnelling protocol is employed by the GPRS support nodes within the same network to transfer user data and other communication related information such as signalling messages. The Base station and the Mobile Station communicate via the air interface specified for GSM/GPRS networks. Packet data channel or PDCH is the term used to describe the physical layer of the GPRS implementation. “Different packet data logical channels can camp on the same PDCH, which are packet data traffic channels (PDTCHs) used for data transfer, packet common control channels (PCCCHs) used to convey the GPRS common control signalling, and packet dedicated control channels (PDCCHs) used to convey the GPRS control signalling for a dedicated MS.”[ 2]
3. Channel Allocation in GPRS/GSM Channel Allocation here is quite elastic as many channels can be allocated to a single user if the need arises or one channel can be shared by several. Every communication medium has a limited number of channels, which makes the issue of managing and allocating of this resource very critical. Efficient usage of available channels is therefore a high priority task in the area of resource management. Past studies were almost totally devoted to this issue. For e.g. A. Xhafa and O. K. Tonguz in their paper “Dynamic priority queuing of handoff requests in PCS” discussed the dynamic priority queuing for voice over GSM in Personal Communication System. J. Wang, Q. A. Zeng, and D. P. Agrawal in their paper “Performance analysis of preemptive handoff scheme for integrated wireless mobile networks” discuss various channel allocation schemes for transmission of multimedia content in integrated networks such as GPRS/GSM. In Integrated networks for voice and data such as GSM circuit switching is employed which hugely beneficial in voice transmission is very inefficient for data transmission due to the nature of data which can have erratic and burst nature. The channel allocation schemes used in such networks do not distinguish between differing bandwidth requirements for voice and data traffic. Both these services are allocated equal slot numbers. GPRS is employed to counter this and handle the data traffic as it employs packet switching. It allows for flexible and on demand based allocation of slot numbers. Various channel allocation schemes have been proposed in the past schemes have been proposed that we now mention. [2] proposed a channel allocation scheme based on the Markov-Modulated Poisson Process (MMPP). [5] analyzed and compared performances of many channel allocation schemes and also put forth a dynamic channel allocation scheme based on queuing. [6] studied the
GSM/GPRS for a Low-Earth-Orbit (LEO) satellite system. [7] proposed two algorithms for channel allocation schemes, i.e., First Fit (FF) and Best Fit (BF ). Marques and Bonatti [8] have proposed the link dimensioning using circuit-switched voice and packet-switched data, while considering that each PDU could use only a single channel at time. They obtain the wireless link dimensioning using an approximation technique that allows the decomposition of the two-dimensional Markov chain, using 2 one-dimensional Markov chains.
4. Priority and Sharing Models The wireless link dimensioning strategy adopted by Marques and Bonatti with voice and data integration has many Quality-of-Service parameters like voice blocking probability, the packet loss probability and the average data and voice packet transfer times. Two different approaches were presented in the paper. In the first, the voice services were taken circuit-switched and the data services packet-switched. The data packets can exclusively use a part of system capacity and share the remaining channels with the voice traffic. In the second both voice and data services are packet-switched, but voice packets having priority over the data packets.
4.1 Sharing In sharing, the voice traffic is circuit switched and the data traffic is packet switched. The channel allocation scheme employed in this model uses a partial method. Here C channels are exclusively reserved for use of data traffic while the rest of the channels Cv are shared for both voic e and data.
Fig. 2 Figure showing C + Cv channels. [6]
When there is a voice call taking place, all the channels remain unavailable for use. At t he end of the conversation they are made available again. When the channel is being used for the transmission of
a data packet it is unavailable unless a pre-emption takes place due to a voice call. When this happens the remaining bytes of the packet are placed in a buffer and assigned a priority tag and sent as soon as the channel is again made available for transmission. Each station has a buffer to store packets and data during such instance. If the buffer is full the packet is dropped. The number of available channels for data traffic is (C + x) with the overall rate of (C + x)γ , where γ is the channel transmission rate. Hence each MS generates packets having Poisson distribution with average total rate λ. E{Z} is the average size of the packet, and average service time is 1/µ . The voice calls are generated within average total rate λv having Poisson distribution, having 1/µv as the average service time. The queue M/M/Cv/Cv is used to describe the no. of channels taken up by voice traffic. The average packet transfer rate here is given by Little’s theorem as given below. [6]
(1)
Figure 3 shows data packet number probability in the system for many ν/µv values.
Fig. 3 Data packet number probability for: ρv =13 Erl, r =13 Erl, and b =50. Conditional model is denoted by a solid line and the Bidimensional Markov chain is denoted by circles. [6]
4.2 Priority In the priority model, Voice and data services share a packet -switching server where the transmission rate is (C + Cv)γ and has an infinite buffer. There is now however a priority for voice packets so as to ensure a small delay for them. The packets that have the same priority obey the First-In, First-Out (FIFO) rule to avoid conflict. The packet transfer rate here is given by the eqn. [6]
(2)
4.3 Calculated Results Using set system values Marques and Bonatti derived their observations for both models as observed in conclusion. For data traffic, r =10 Erl and ρv =10 Erl
Fig. 4 Average data transfer time for data traffic. [6]
For voice traffic, r =10 Erl and ρv =10 Erl
Fig. 5 Average transfer time for voice traffic. [6]
5. Conclusion Due to several numerical experiments some remarks can be made out of the paper by Marques and Bonatti. There are several tradeoffs when wireless link dimensioning is used. This is due to the poor quality estimate of the traffic quality. Minimal capacities satisfying QoS parameters and the system sensibility to traffic overloads give results that are inversely proportional to each other. There is a small gain in system capacity on using the channel sharing technique when compared to the non-sharing one. On using the priority model the number of channels to satisfy the voice blocking probability obtained were enough to accommodate the low data traffic. When the voice traffic is much bigger than the data traff ic it was found that the capacities determined with the priority model were smaller when compared to the sharing model. When the voice traffic is bigger than the data traffic the capacities determined with the priority model is smaller when compared to the sharing model. The computed average data packet transfer time sensibility to voice or data traffic overload of the nominal capacities with the priority model is smaller than sharing model. Both models have their place in different scenarios and both are quite effective.
6. References 1. Yan Zhang; Boon-Hee Soong; , "Performance evaluation of GSM/GPRS networks with channel re-allocation scheme," Communications Letters, IEEE , vol.8, no.5, pp. 280- 282, May 2004 2. Phone Lin; Yi-Bing Lin; , "Channel allocation for GPRS," Vehicular Technology, IEEE Transactions on , vol.50, no.2, pp.375-387, Mar 2001 Y.-B. Lin and I. Chlamtac, Mobile Network Protocols and Services. New York: Wiley, 2000 3. 4. C. Lindemann and A. Thummler, “Performance analysis of the general packet radio service,” in Proc. IEEE Distributed Computing Systems, pp. 673–680, 2001. 4. A. G. Qureshi and A. S. Ween, “Dynamic resource allocation for GSM-GPRS services over a LEO satellite system,” in Proc. IEEE ICICS ’97, pp.20–24, 1997. 5. A. Tripathi and K. N. Sivarajan, “Resource allocation in GPRS wir eless networks,” in Proc. IEEE ICPWC ’00, pp. 388-394, 2000. 6. de Oliveira Marques, M.; Bonatti, I.S.; , "Wireless link dimensioning: priority versus sharing," Telecommunications, 2005. advanced industrial conference on telecommunications/service assurance with partial and intermittent resources conference/e-learning on telecommunications workshop. aict/sapir/elete 2005. proceedings , vol., no., pp. 135- 139, 17-20 July 2005
