1. The general scenario
Consider a university with 2 campuses A and B. Campus A is the campus where all central IT facilities (PABX, VoIP server, Intranet server) are located. Campuses are connected with each other and with the outside world as shown in the figure below.
The central servers include, depending on the scenario, PABX, Intranet www and file servers, mail server etc.
The I/F denotes interface equipment to the connecting link or circuit group. Depending on the scenario, it might be a PDH multiplex interface, IP router, etc.
Common assumptions:
1. There are 2000 staff and 4000 students in Campus A, and 1000 staff and 4000 students in Campus B.
2. Assume that all telephone and data traffic generated by the staff and students can be considered Poisson. This is a crude assumption (please check the slides on Internet traffic), but necessary for simplicity.
3. Only the staff members generate phone calls. Assume that all staff generate exactly the same telephone traffic. The traffic is distributed uniformly among staff members, i.e. staff member X sends/receives the same amount of telephone traffic to/from staff member Y as they do to/from another staff member Z.
4. Every staff member has a telephone on the desk and a computer (hence they generate both telephone and data traffic). Assume that all staff members generate the same amount of data traffic.
5. Students generate only data traffic. Assume that all students generate the same data traffic.
Task 1:
Make realistic assumptions as to the traffic generation behaviour of an individual staff member and individual student. In other words, assign numbers to the following:
a) Average arrival rate of telephone calls to/from each staff member's telephone λi (assume that this includes both outgoing and incoming calls and that λi will have the same value for all staff ).
b) Average call holding time (assume the same value for all calls).
c) Proportion of internal telephone traffic i.e. proportion of telephone traffic exchanged with other staff (this will be uniformly distributed across all staff).
d) The average downlink data traffic (file and email downloads, web pages) per staff member in bps (or kbps or Mbps - whatever suits the purpose).
e) The average downlink data traffic per student.
f) Proportion of downlink data traffic originating at organisation's internal servers (note: the remaining downlink data traffic will originate at external sources). Assume that this proportion is the same for staff as it is for students.
g) The average uplink data traffic (web, email etc. requests, TCP ACKs, uploads of emails) per staff member in bps (or kbps or Mbps).
h) The average uplink data traffic (web, email etc. requests, TCP ACKs, uploads of emails) per student in bps (or kbps or Mbps).
i) Proportion of uplink data traffic terminating at internal servers (the remaining part will go to the outside world). Assume that this proportion is the same for staff and students.
Provide rationale for each assumption (i.e. explain why you consider it realistic).
Task 2:
For this task, only consider telephone traffic generated by staff members. Assume that all telephone traffic is circuit-switched as in the traditional PSTN. The relevant scenario is illustrated in the figure below. The "central servers" become a PABX, the "interface equipment" at Campus B becomes a telephone line concentrator with PDH multiplex equipment, and the interface at Campus A becomes PDH multiplex equipment, as in the figure.
a) Given the assumptions made in Task 1 a,b,c, map the total telephone traffic flows (in Erlangs) onto circuit group A-B and the external circuit group. Show the reasoning leading to the answer you gave.
b) If we require that the end-to-end GoS probability of loss for each call (internal and external) is no greater than 0.01 (1 in 100), what are the GoS probabilities of loss allowed for each of the two groups of circuits?
c) Given the results in a and b above, what are the required numbers of circuits in each group?
Task 3:
For this task, only consider data traffic as specified in Task 1 d,e,f,g,h,i. As illustrated in the figure below, the "central servers" now become the email/www servers, and the "interface equipment" is simply IP routers.
a) Map the total flows of data (in kbps or Mbps) generated by staff and students onto the links connecting Campus B with A and Campus A with outside world. Please assume that both links are part of WAN (leased from external providers), not part of the organisation's Ethernet. As a consequence, for each link you have to state separately the downlink and uplink traffic. Provide rationale (explanation) for your answer.
b) Given the traffic flows obtained in a above, state the data rate capacity (in Mbps) required for each link (separately in the downlink and uplink directions), under the condition that average packet delay for each link cannot exceed 2 ms. Assume that the internal data transfers between servers, hosts and routers located in the same Campus incur negligible delay, and that the processing delay at servers and routers is also negligible. Note: you may need to make a realistic assumption regarding the average packet length (why?)
c) Calculate the buffer sizes required on the link A→B and the link B→A to ensure that the packet loss in no more than 1 in 2000.
Task 4:
For this task, assume that all telephony needs of the organisation are served with the VoIP solution. Assume that the VoIP codec rate is 64 kbps. Assume that the modem transmits data only when there is a talkspurt, and that it does not transmit anything during the silence period. Assume the average talkspurt duration of 0.6 s and the average silence duration of 0.4 s, and that in a typical phone call 50% of talkspurts are transmitted in one direction, and 50% in the other (i.e. that either one or the other party in the call can talk and that they talk, on average, for the same amount of time). This results in the symmetry of downlink and uplink traffic loads incurred by the VoIP call.
a) Calculate the additional VoIP data traffic (in kbps or Mbps) loading each of the two links (external, A-B) in each (uplink and downlink) direction. Note: to find out the average data traffic generated by a VoIP call in each direction, you may reuse the calculation of data source as given in the tutorial materials, but in the case at hand there are much simpler ways to get the same result. Please try to keep it simple! I will not give additional points for data source calculation - in fact, I may take them away
b) Given that all traffic in the organisation is now data traffic, calculate the total data traffic for each link in each direction (i.e. add the VoIP traffic to the data traffic resulting from Task 3a).
c) Will the addition of VoIP traffic to the remaining data traffic make significant difference to the dimensioning of the data links? Justify your answer.