WS1F11 - Worksheet#1 EE122 Fall 2011 You are not required...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Worksheet #1 EE122 Fall 2011 You are not required to do these problems, and we are not asking you to hand them in. However, if you can't do these problems, then you will have trouble on the final. In fact, many of these questions came from last year's final. As the saying goes, "Know This". 1 1. Acronyms Consider the following acronyms: A, AIMD, BGP, CDN, CIDR, CNAME, DHCP, DNS, ECN, FIN, HTTP, ICANN, LPM, MSS, MTU, MX, NAT, NS, PMTU, PTR, RFC, RIP, RST, RTO, SSTHRESH, STFU, SYN, TCP, TLD, UDP. Match the acronyms to the descriptions below, using each acronym exactly once. Supply exactly one answer for each item below. In some cases, an acronym might fit more than one question, but there is only one way to answer these so that every acronym is used exactly once. For terms that aren't familiar to you, use the book or the web. i) A mechanism that allows the network to indicate congestion to TCP without having to drop the packet. ii) The DNS record type that provides the IP address of a given hostname. iii) The DNS record type that provides the canonical name associated with a given hostnam. iv) The DNS record type that provides the authoritative name server associated with a given domain. v) The DNS record type that provides the mail server associated with a given domain. vi) The DNS record type that provides a hostname associated with a given IP address vii) The name of the message used to initiate a TCP connection. viii)A message (control flag) used to terminate a TCP connection abruptly. ix) A message (control flag) used to terminate a TCP connection smoothly. x) The kind of window adjustment algorithm used in TCP. xi) The size of the largest IP packet that can be sent end-to-end along a path without fragmentation. xii) The largest sized segment that can be sent by TCP. xiii)The largest size IP packet that can be sent across a particular link. xiv) The most common examples of this acronym are .com, .net, and .org. 2 xv) The documents used by the IETF to describe protocol standards. xvi) An in-line network device that rewrites IP addresses (and often transport ports). xvii) A widely used routing protocol that does not necessarily compute lowest- cost paths. xviii) The organization that is ultimately responsible for address allocation and domain name management. xix) A method for matching packets to routing entries that determines which among multiple matching entries determines the packet's destination. xx) The length of time TCP waits for an acknowledgement before timing out and initiating a retransmission. xxi) The most widely used reliable transport protocol. xxii) An unreliable transport protocol. xxiii) The protocol that provides newly arrived hosts with their own IP addresses (and also gives them other information). xxiv) An IP addressing scheme that does not involve the traditional A, B, and C address classes. xxv) Akamai is an example of one. xxvi) The TCP congestion control state variable that determines when to switch from slow-start to congestion avoidance. xxvii) An application-level protocol that typically runs over TCP. xxviii) An application-level protocol that typically runs over UDP. xxix) A distance-vector routing protocol. xxx) The appropriate response to Stanford students when they bring up the 2010 Big Game. 3 2. A TCP Sob Story Due to budget cuts, the CS department is forced to buy cheaper networking equipment. It turns out that one of the switches has a defective port that has a strange dropping behavior. On a particular port, it drops every fourth packet, in both the sending and receiving directions. To be precise, if one only looks at the packets being received, it drops the 4th, 8th, 12th, ... packet. Similarly, if one only looks at the packets being sent, it drops the 4th, 8th, 12th, etc. This problematic port is the port that attaches the department chair's computer, host A, to the network, so every fourth packet sent to A is dropped, and every fourth packet sent by A is dropped. Suspecting a problem, the CS department runs a test by initiating a file transfer from host A to an offsite host B. The RTT of the packets is 1 msec, the packet transmission time is .01msec, and the RTO (timeout value) in the TCP connection is 100msec (you won't need these exact numbers, they just set the relative magnitude of the various timescales). Host A uses a transport protocol with a sliding window flow control with a constant window size of 5 packets and a duplicate ACK threshold of 3 packets (3 duplicate ACKs lead to a retransmission). The test consists of sending a series of data packets (labeled D1, D2, ...) from A to B. We want to describe the transmissions from A (by writing Dx for the appropriate x). The first five data packets are shown, in addition to the starting SYN and ACK packets, just to get you started (and just to clarify, we assume that the packet count, in terms of dropping, starts with the SYN; therefore, in these first few packets, D2 and only D2 is dropped). The packets from A that will be dropped are marked in bold. Please fill in the remaining 15 packets sent from A. (And remember that the packets coming to A are also dropped in a similar one-every-four pattern, with the packet count starting with the SYN-ACK, so the first packet dropped in the receiving direction is the ACK sent in response to D4.) SYN, ACK, D1, D2, D3, D4, D5, ____, ____, ____, ____, ____, ____, ____, ____, ____, ____, ____, ____, ____, ____, ____ How many fast retransmits occur during the transmission of this set of packets?_____ How may timeouts occur during the transmission of this set of packets?______ 4 3. TCP Congestion Control Details In the following, assume that the MSS=1000 bytes and that in all computations CWND is rounded down to the nearest integer. During the congestion avoidance phase, TCP updates the window following each ACK by the following equation: CWND += MSS / Int(CWND / MSS) Where Int(x) is the integer part of x. In the slow-start phase, TCP updates the window following each ACK by CWND += MSS The algorithm leaves slow-start when CWND > SSThresh (note the strict inequality, which is implementation dependent rather than a property of the TCP spec.). When undergoing a Fast Retransmit, assume the TCP uses the "advanced" version discussed in the congestion control lecture (slide 38). In the answers on the next page, denote the packet containing bytes 1401 to 1500 as the 15th MSS, and denote the ACK expecting the next byte of 15001 as an ACK for the 16th MSS. Consider a TCP connection with a CWND of 12000 bytes and then undergoes a timeout. Assume all data up to 16000 have been ACKed (i.e., the last ACK's expected next byte was 16001). What is the value of CWIN, and which packets does TCP send, at each step in the following process? The first two lines are filled out, to indicate the required form of the answers. 5 ACK received or timeout Timeout cwin 1000 2000 ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ ________ Packets sent (MSS #'s) 17 18, 19 _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ _________ ACK for 17th MSS ACK for 18th MSS: ACK for 19th MSS: ACK for 20th MSS: ACK for 21st MSS: ACK for 22nd packet: ACK for 23rd packet: ACK for 24th packet: ACK for 25th packet: ACK for 26th packet: ACK for 27th packet: ACK for 28 th packet? ACK for 29th packet: ACK for 30th packet: ACK for 31st packet: ACK for 31st packet? ACK for 31st packet: ACK for 31st packet: ACK for 31st packet: ACK for 31st packet? ACK for 31st packet: ACK for 31st packet: ACK for 39th packet? 6 4. HTTP and Caching i) Consider a case where a client A is retrieving files F and G from web site B. F and G are both 125KB (i.e., one megabit). The RTT between A and B is 10msec (note, these are round-trip-times, not one-way latencies), and the bandwidth between the sites is 10Mbps. Assume all TCP SYN/ACK packets and HTTP request packets are negligible in size. How long does it take A to retrieve both files under the following circumstances: Sequential (one-at-a-time) requests with nonpersistent TCP connections:_____ Concurrent requests with nonpersistent TCP connections:_____ Sequential requests within a single persistent TCP connection:_____ Pipelined requests within a single persistent TCP connection:_____ ii) Consider the same situation as in i), but assume that rather than a dedicated link there is a large shared link with many flows traversing it, and each TCP connection gets 10Mbps (adding additional flows does not significantly change the bandwidth per TCP connection, because there are thousands of flows on the link). Now, how long does it take A to retrieve both files under the following circumstances: Sequential (one-at-a-time) requests with nonpersistent TCP connections:_____ Concurrent requests with nonpersistent TCP connections:_____ Sequential requests within a single persistent TCP connection:_____ Pipelined requests within a single persistent TCP connection:_____ iii) Consider the same situation as in i), except that A is only downloading file F and there is now a cache C between A and B. All requests from A to B go through cache C, and assume the bandwidth along the path from A to C is 1gbps and the RTT between A and C is negligible, while the bandwidth along the path from C to B is 10mbps with an RTT of 10msec. Note, these are round-trip-times, not one-way latencies. As above, assume that the file is 125KB (i.e., one megabit) and that all TCP SYN/ACK packets and HTTP request packets are negligible in size. 7 Assume the cache operates as follows: (where the origin server refers to the site named in the URL) If the object is not in the cache, the request is forwarded to the origin server If the object is in the cache, and the cache entry has not timed out (i.e., the cache TTL has not expired), the object is returned to the client If the object is in the cache, but the cache entry has timed out, the cache issues a conditional-GET to the origin server, asking if the object has changed since this object was cached: if the origin server responds that it hasn't, the cache returns the cached object, otherwise the origin server responds with the updated object which the cache forwards to the client. How long does it take for A to receive the file under the following circumstances: The file is not in the cache._____________ The file is in the cache and the TTL has not expired. _____________ The file is in the cache, the TTL has expired, but the file has not been changed. _____________ The file is in the cache, the TTL has expired, and the file has changed. _____________ 8 5. BGP Policy Extravaganza i) Length-based policies: A D E F B C H G Consider the interdomain connectivity graph above, where each of the ASes A through H are running BGP. Every domain advertises its chosen routes to all its neighbors (i.e., there is no limiting export policy). Assume that in this case all the domains want shortest paths (as measured in terms of the number of ASes traversed) and, when there is a tie, they prefer to route advertised with the letter closest to the beginning of the alphabet. What path do packets take from A to G?__________ What path do packets take from G to A?__________ ii) Export policies: Using the same graph, now assume that nodes only advertise routes to the vertically connected node and a horizontal peer if they have one: assume that A-D are peers, E-F are peers, and C-H are peers. When advertising routes, nodes either advertise all or none. When selecting a route, ASes prefer the shortest path and, when there is a tie, they prefer to route advertised with the letter closest to the beginning of the alphabet. What path do packets take from A to G?__________ What path do packets take from G to A?__________ iii) Nonstandard policies: Using the same graph, now assume that each domain's route preferences can be expressed in terms of their preferences over their neighbors (i.e., a domain prefers any route coming from one neighbor over any route coming from another neighbor, no matter what their relative lengths are). Assume the preferences are as follows (where x > y means the domain prefers routes from x over routes from y). Domain A: D>B Domain B: A>C Domain C: H>B>D Domain D: C>E>A Domain E: H>F>D Domain F: E>G 9 Domain G: F>H Domain H: C>E>G What path do packets take from A to G?__________ What path do packets take from G to A?__________ Note: there is not a unique right answer to this question: BGP could reach several different fixed points. All you need to do is give an answer that is stable; that is, your answer should reflect a set of routes that, if BGP were to start with these routes as its initial condition, BGP would continue to retain that set of routes. 10 ...
View Full Document

This note was uploaded on 01/09/2012 for the course ELECTRICAL 122 taught by Professor Shenker during the Spring '11 term at Berkeley.

Ask a homework question - tutors are online