S-38.188 exam, 16.5.2005

General comments about the grading

• correct acronym gives 0.5p and a short explanation 0.5p
• some minor errors were tolerated in the acronyms (for example, one of the letters of the acronym could be false without penalty), as long as the explanation was correct

• OSPF acronym 0.5p, routing inside AS 0.5p
• correct explanation of flooding + topology + Dijkstra 1.5p
• other descriptions of the operation: updates 0.5p, cost metrics 0.5p, memory requirements heavier than in RIP (why?) 0.5p, stability issues 0.5p, security aspects/routing areas 0.5p
• maximum was still 4p

• MAC addresses: 48 bits, unique, non-hierarchical, hard-coded = 1p
• IP addresses: 32 bit or 128 in IPv6, usually unique = 0.5p; hierarchical (important!)= 1p
• CIDR: reason + explanation, CIDR address format = 1p
• transport layer: IP address + 16 bit port number = 0.5p
• application layer: for example SMTP: user@domain or HTTP: URL ("domain name + directory path" ) = 0.5p
• maximum was still 4p

• UDP: acronym (0.5p), unreliable (0.5p), used mostly by real time applications (0.5p)
• TCP: acronym (0.5p), connection oriented (0.5p), reliable (0.5p) congestion cntrl (0.5p), flow cntrl (0.5p), applications (0.5p)
• extra pluses were given for discussions related to fairness (greedy vs. TCP friendly) and addressing (ports and their use for muxing)
• maximum was still 4p

The answer is on the first slides of the network management lecture. FCAPS, 0.5p per area, 0.5p per explanation. As you can notice, this sums up to five points only, and thus one point is reserved for general discussion, e.g. on the overlap of certain functions for implementing management of different areas. There are some people who got the FCAPS model right but did not try to discuss it, and some people who discussed it but forgot some functional areas.

Some answers used other types of classifications of network management tasks and I tried to check if the tasks corresponding to the different parts of the FCAPS model were present. Also, an extra point was given to some clever classification.

The main problem on this question for many was the mix-up of network management with control mechanisms implemented in the network. This differentiation can be done for example by thinking of time scales -- the automated mechanisms such as congestion control act in very short time scales, while in network management the time scale focus is from minutes to even days or longer. Thus, discussion on flow and congestion control or ICMP did not yield points.

Some people who actually got to the problem domain (i.e. network management) were stuck in details such as MIBs and SNMP, and forgot the big picture which was the topic of the question. Naturally, SNMP can be used to implement part of configuration management, but it is just a tool. This was especially unlucky since there were a couple of brilliant answers to a question of what SNMP and MIB are, but since this was not the question I could not give points.

Problem 6: Real time services

• Issues related to real time transport
  • general issues: delay (0.5p) , delay variation (0.5p) and loss (0.5p)
  • mechanisms at application layer: buffering (0.5p) and rate adaptation (0.5p)
  • different needs of interactive/streaming applications (1p)
  • QoS support at network layer (DiffServ/IntServ) 0.5p
  • maximum 3p for this part
• RTP/RTCP
  • acronyms (0.5p for knowing both)
  • Functions of the RTP/RTCP protocols: provides timing info, synchronization of multiple streams, negotiation of encoding, congestion info (reports packet losses) 2p
  • RTP for transmission of packets and RTCP is the control part (periodic control messages) (1p)
  • maximum 3p for this part
  • many mixed up RTCP and RTSP - RTSP is for controlling streamed media (pause, stop, rewind functions)

Valitustilaisuus

A "valitustilaisuus" will be arranged for questions about the grading. The time and place is:

• Thursday, June 23, at 10 o'clock, room D302.

Results

The grading thresholds for the exam are at the beginning of the table below.