Transcript
p1
EXAM A TSM 350 – IP Telephony MY NAME IS _______________________________________ Winter 2008 Final Exam: April 24, 2008 Directions: • The exam is closed book and notes. All specific data will be provided. • Please answer all questions (multiple choice/short answer/True-False) on the exam page. Please provide ladder diagrams on the reverse side of the exam page. Use the blue books for scratch; edit your answers on the blue books and only transfer to the exam page when you are done!. • Please keep the short answers SHORT and answer only in the space provided. We will ignore the portion of your answers that do not fit into the provided space. Again, use the blue books for editing. • The exam will be graded, and solutions posted, on Friday April 25. Final grades will be passed to the registrar on Friday as well. You can pick up the graded final exams from Prof. O’Reilly beginning Monday April 28, as I will be on vacation until May 12. • Lets try to limit the time to 2.5 hours. • Best of luck to all of you. Please stay in touch (particularly if/when you form the next $4B telecom startup, and you are looking for folks to sit on your board of advisors). 1.
You are providing specifications to carriers to provide PSTN termination of traffic from a VoIP PBX. On the VoIP side, you expect to use 30 msec frames of G722.2. The raw bit rate for G722.2 is 24 kbits/sec and the delay introduced by either the G722.2 encoding or decoding is 6msec (to be more clear, 6 msec for encoding and 6 msec for decoding). There is very little jitter on your network, and so the jitter buffer at the VoIP gateway is set for 1 frame. The maximum transit on the combination of VoIP network and PSTN network is expected to always be less than 82 msec (e.g., the transit time for each call will be between 0 and 82 msec). A. What is the IPv4 packet size (layer2&3+data) and the wire data rate (kbits/sec) for carrying the VoIP RTP on the customer’s etherenet LAN (just show results – put careful and clear calculations on the back of this page using at most ½ page) (10 pts): 30 ms * 24 kb/s = 720 bits or 90 bytes. Add that to 20(ip) + 12(rtp)+8(udp)+ 28 (Ethernet) and you get 158 bytes in 30 msec, or 42.something kbits/sec. B. What is the maximum relative echo level [dB] that can be tolerated for some calls and still achieve acceptable voice quality? (5 pts) max acceptable echo level when the delay is 0 msec is (from graph) about -21 dB. C. What is the maximum relative echo level [dB] that can be tolerated so that ALL calls will have an acceptable voice quality? (5 pts). The max echo for all calls is with max delay, which is [msec] 82 (transit) + 30 (packetization) + 6 (codec encode) + 6 (codec decode) + 30 (jitter buffer) = 154 msec. This equates to about a -45 dB relative echo (reflection). D. Is this call likely to be considered a domestic toll quality call? TRUE or FALSE (5 pts) . (False – domestic toll quality is something like 100 or 120 msec of delay).
2.
In a Femtocell access point, the radio side of the access point is likely to belong to one of the following protocol families: (5 pts) A – CDMA F – (A), (B), (C) B – WiFi G – (B) and (D) C – GSM H – (A) and (C) D – Bluetooth I – (A), (B), (C), and (D) E – DECTThe radio access point for a Femtocell is going to be either CDMA or GSM, so the answer is H
p2
3.
In a Femtocell access point, the network (towards the carrier) protocol is likely to be: (5 pts) A. Circuit switched (either DS0s or T1/PRI) B. Some form of VoIP/SIP or wireless protocol encapsulated in IP C. Negotiated at connection time with the wireless carrier D. Up to the customer to specify, since it’s riding on the customer’s network. B: The network protocol is going to be a wireless protocol or VoIP/SIP encapsulated in IP.
4.
Provide one reason why it is sometime useful to re-direct a call from one sip user agent to another (less than or equal to one sentence that fits on the lines below .. try your answer before writing!) (5 pts) Call forwarding Call Transfers Third Party Call Control
5.
You have been asked to implement an Asterisk based VoIP PBX for a client. The asterisk server and the phones all share the same ethernet subnet. The client wants to monitor all calls, and thus is asking you to force all the RTP traffic to go through the Asterisk server. Compared to a scenario where the RTP goes directly from VoIP phone to VoIP phone, this results in: (5pts) A. Doubling the traffic through the subnet switch B. Halfling the traffic through the subnet switch C. Doesn’t matter; the same amount of traffic will be running, just in different places. A - Doubles the amount of traffic through the switch. . If calling UA is A, the called UA is B, the asterisk server is C, and the “Switch” is S, then without the asterisk server the traffic goes A => S => B. With the asterisk server the traffic goes A => S => C => S => B. RTP packets traverse the switch twice.
6.
Your local VoIP carrier is interested in increasing the frame (sample buffer) size from 20 ms to 40 ms in their VoIP gateway, which converts telephone calls from the PSTN to IP packets and which uses the G729.r8 codec. If they do this, it will (5 pts): A - increase the efficiency of carrying VoIP traffic, but also increase the end-to-end delay and vulnerability to packet loss. B - increase the efficiency of carrying VoIP traffic and make voice quality less sensitive to packet loss but at the expense of increased end-to-end delay, C - increase the network latency and improve bandwidth efficiency and QoS. D - increase .latency with no effect on QoS or bandwidth efficiency. E – A and D F – A and G G – decrease the average bandwidth per call Answer is A. B is wrong because the change makes the call more sensitive to packet loss. C is not a good answer because it’s hard to say anything about QoS, could be better could be worse. D is wrong because this will increase the bandwidth efficiency (same reason G is wrong).
7. In many VoIP systems, the first invite from user-agent to registration server (or proxy) is generally rejected (with a SIP/2.0 401 Unauthorized or a SIP/2.0 407 Proxy Authentication Required message). Why wouldn’t the user agent pass an encrypted version of the sip password in the first message sent to the server, and thus avoid having to send two sip messages? (10 pts) (answer must fit into the space below and be readable. Edit and re-write your answer before entering it below!). VoIP typically uses digest authentication to authenticate an endpoint for registration (getting calls) and invites (sending calls). When the proxy rejects the first invite, it includes a nonce as part of the
p3
response, which is hashed along with the sip passwd and provides a one-time secure key back to the server in the second invite. Without the nonce, the encrypted response would be static, inviting replay attacks in which a lurking bad guy could simply sniff and provide the response. 8. ENUM provides IP to PSTN number conversion, so that VoIP traffic can flow through the PSTN: True or False? (5 pts) FALSE – ENUM provides PSTN to IP number conversion so that traffic can AVOID the PSTN.
9. The following URLs allow calls to be made without access to the PSTN: (2 pts each) a. sip:
[email protected] True or False b.tel:
[email protected] True or False True or False c. sip:
[email protected] These are all true, since they provide a host name or ip address to direct traffic to. 10. The tone of this class reminded me of Dostoevsky’s The Brothers Karamazov, in its bleak depiction of life as a struggle between faith, reason, and free will. (1 pt) True or False . True or False. Everyone should try to get through The Brothers Karamazov once in their life. But if you can actually finish it without going on medication for depression, you’re a better person than I. Definitely not recommended for graduate students – wait until you are older and more cynical! For graduate students, I recommend the wonderful British SciFi writer Ian Banks. Try Player of Games from his Culture series – GREAT STUFF. 11. What is ngrep, why is it useful? (10 pts) (2 sentences, answer must fit in space below. Think, work on your answer in the blue book, and then write it down). (5 pts) ngrep (network grep) is a unix utility that provides a text based (command line driven) dump and protocol analysis of network traffic. You can think of it as the poor person’s wireshark – protocol analyzier. Because it’s command line drive with text output, it’s very light weight and easy to use remotely compared to something like wireshark which has a x windows based client (UNIX) or Microsoft Windows based client. 12. On the back page of this exam (start on p1) draw a ladder diagram of the call described by the attached sip debugs(20 points). Use the blue book for scrap and don’t copy onto the back page until you are sure it’s correct In ONE SENTENCE explain what’s going on: (10 points), and how it differs from a similar call flow that you did in the second in class lab. This is NOT a trick question. The call is a blind call transfer. In the lab we did a consultative call transfer. A blind call transfer is one in which the calling party is immediately transferred to the transferred party without there being a step in between (e.g., with out the called party talking to the transferred party).
p4
2009 129.10.39.222
Astetisk 129.10.34.100
2001 129.10.39.219
Invite:2001 & SDP 407 P.A.Req. ACK Invite:2001 & SDP Trying Invite:2001 & SDP Trying Ringing Ringing 200 OK & SDP ACK 200 OK & SDP ACK RTP Invite & SDP Trying 200 OK & SDP Invite & SDP 200 OK & SDP ACK RTP ACK Refer 202 Accepted Notify Notify 200 OK 200 OK BYE 200 OK Invite:2003 & SDP Trying Ringing Invite & SDP 200 OK ACK
2003 129.10.39.217
