hwk1_570_24s_sol

.pdf

School

University of North Texas *

*We aren’t endorsed by this school

Course

3530

Subject

Electrical Engineering

Date

Feb 20, 2024

Type

pdf

Pages

Uploaded by BailiffPorcupine4131

1 Problem 1 Short Questions (27 pts; 3pts each): Your answers should be precise-and-concise. A sentence or two of an answer is sufficient for most of these questions. A. Based on definitions of the edge and the core networks, identify which category does each of the following networks belong to: a. An IXP network: core b. NCSU campus WiFi: edge c. The cellular network that your T-Mobile/Verizon mobile phone directly connects to: edge d. Facebook’s wide area network that interconnects its multiple data centers: core e. The inter-continent cable under the Atlantic Ocean: core f. A road side unit which an autonomous driving vehicle is connected to: edge B. Explain what a Point-of-Presence is? It’s a set of routers / a subnetwork of the internal network of an ISP that connects to its external peering networks. C. Explain two reasons why packet switching is best suited for the Internet than circuit switching? Packet switching 1) is great for bursty data, 2) enables resource sharing without wastage (or can serve more users), 3) does not require circuit setup which incurs setup overhead, and 4) is more resilient to failures. D. Assume a packet is sent from a sender and is to be delivered to a remote destination through multiple intermediate switches or routers. The receiver however does not receive the packet in the end. What could have happened? The packet may have gone through a switch/router with insufficient buffer / memory, and hence got dropped at that switch/router. E. Are buffers / queues needed for a fully circuit switching network? No. Because resource is dedicatedly reserved in circuit switching. So as long as a user is not sending at a speed more than the reserved bandwidth, queueing will not happen. F. You went into a room with an Amazon Echo (or any other device with a voice-controlled digital personal assistant (DPA) similar to Alexa), and say “Alexa, what’s the time?” Alexa replies: “The time is 1:59pm.” Describe the above process in the realm of a protocol between you and Alexa. (Consider the three components of a protocol, and explain what each component is in this scenario.) Syntax: [Name, or Alexa] + [Query to Execute] Semantics: A sender sending the voice message asks the DPA with name [Name] to execute the [Query to Execute]. Action: Upon receiving the voice message, the DPA with [Name] executes the [Query to Execute], and plays the response as an audio message back to the sender. G. Explain how layering is implemented when a sender is sending out a packet.

2 Each layer (beside application layer) in the sender’s network stack will add a packet header, which contains the necessary information needed for the same layer at a remote network device or end host to process the packet and deliver it to the correct upper layer / next-hop entity. H. How can you distinguish different types of network devices such as a hub, a bridge, a router and a host? The layers of protocols that the device implements. Hub: physical layer only. Bridge: up to data link layer. Router: up to network layer. Host: up to application layer. I. Describe the major difference between the OSI and the TCP/IP layering models, and whether each of the different layers is end-to-end or not. TCP/IP model does not have the presentation and the session layers in the OSI model, which are both merged into the application layer. Both these layers are end-to-end. Problem 2 Protocol (9 pts; 3 pts each): Assume a protocol stack has 5 layers, with the topmost layer being the application layer. The four lower layers each adds an overhead of 30 bytes as headers and trailers, and does not change the packet payload. Consider the application layer sending application PDUs with different sizes as follows. Compute the percentage bit overhead in this stack as seen over a long period of communications. Use the application layer PDU as the basis. A. Application sends 200-byte fixed-length PDUs at all times. Average PDU size = 200 bytes Per-PDU overhead = 30 * 4 = 120 bytes Per-PDU overhead ratio = 120 / 200 = 60% Note: It is also okay to compute based on the final physical layer PDU size in this problem, which is 120/320 = 37.5%. But since application PDUs are the only meaningful data transmitted, the application PDU is commonly used as the “basis” of computation such as delay, throughput, loss ratio, etc. B. Application sends 200-byte PDUs at half of the times, and 400-byte PDUs at the other half. Average PDU size = (200+400)/2 = 300 bytes Per-PDU overhead = 30 * 4 = 120 bytes Per-PDU overhead ratio = 120 / 300 = 40% C. What might be the reason that we do not want to have very large PDU sizes (e.g., 1MB data per packet) to make the overhead of layering as low as possible? There are several reasons. Each one would be sufficient:

3 - The larger a packet, the more likely there will be transmission errors, which could lead to more packet retransmissions. - When a packet is lost due to either packet drop or network failure, recovery would be very costly as the sender needs to retransmit the entire packet again. - A huge buffer is needed to store all packets queuing up at an output port, which could be costly. Problem 3 Packet Loss (8 pts): Assume a client connects to a server via a lossy link. When an application is sending PDUs of different sizes to the server over the path, compute the expected number of bytes lost when transmitting an arbitrary PDU. A. (1pt) Assume the client does not have reliable data transfer. The link has packet loss probability of 20%. Application sends 200-byte fixed-length PDUs at all times. Average PDU size = 200 bytes Expected lost bytes = 200 * 20% = 40 bytes B. (1pt) Assume the client does not have reliable data transfer. The link has packet loss probability of 20%. Application sends 200-byte PDUs at half of the times, and 400-byte PDUs at the other half. Average PDU size = 300 bytes Expected lost bytes = 300 * 20% = 60 bytes C. (1pt) Assume the client does not have reliable data transfer. The link has packet loss probability of 20%. Application sends PDUs following this distribution: 100-byte (10%), 200-byte (30%), 400-byte (50%), 1000-byte (10%). Average PDU size = 370 bytes Expected lost bytes = 370 * 20% = 74 bytes D. (2.5pts) Assume that the client transmits 1 million bytes of data to the server. The PDU goes through a 5-layer network stack with 25 bytes of overhead for each layer except the application layer. During the transmission, exactly 1 bit is corrupted during the transmission, and the corresponding packet must be entirely retransmitted to ensure reliable data transfer. How many bytes in total, including both data and overhead, need to be transmitted, if the PDU size is 100, 1000, 2000, 10000, and 200000 bytes? The 1 million bytes must be transmitted all for once, which does not depend on the PDU size. The overhead depends on the PDU size. Let the PDU size be denoted by p . The number of packets needed to transmit, excluding the retransmission, is 10 6 / p. Each packet has a 4*25=100 byte overhead. Exactly one packet fails, and hence an additional PDU with (p + 100) bytes of data and overhead needs to be retransmitted. So the total (retransmission + overhead) size is: f(p) = (10 6 / p) * 100 + (p + 100). For a PDU size of 100 bytes: (10 6 / 100) * 100 + (100+100) = 1000200 bytes For a PDU size of 1000 bytes: (10 6 / 1000) * 100 + (1000+100) = 101100 bytes For a PDU size of 2000 bytes: (10 6 / 2000) * 100 + (2000+100) = 52100 bytes For a PDU size of 10000 bytes: (10 6 / 10000) * 100 + (10000+100) = 20100 bytes For a PDU size of 200000 bytes: (10 6 / 200000) * 100 + (200000+100) = 200600 bytes

4 E. (2.5pts) What is the optimal PDU size for the above network scenario in D, i.e., requiring the fewest number of bytes to be transmitted in total? Why? Obviously between the PDU sizes given, p = 10000 bytes results in the minimum number of bytes to be transmitted. However, to prove this is indeed the smallest number, we need to show that p=10000 minimizes the function f(p) = 10 6 / p * 100 + (p + 100). To see this, we can take the derivative of f(p) as f’(p) = - 10 6 * 100 / p 2 + 1. The derivative f’(p) is equals to 0 when p = 10000. You can further verify that when p < 10000, f’(p) < 0; when p > 10000, f’(p) > 0; hence p = 10000 indeed corresponds to the minimum point of f(p), proving the result. Problem 4 Transmission Delay (31 pts): Consider the figure below, with three links, each with the specified transmission rate and link length. Assume the speed of light propagation delay on each link is v = ࠵? ∗ ࠵?࠵? ࠵? m/sec . A network layer PDU of length P = 1000 Bytes is to be transmitted from A to B starting from time 0. The PDU originates from the network layer at A, and is considered received when being processed by the network layer at B. Between the network and the DLC layers, the communications happen over a shared memory with read/write speed of 1 Gbps on all nodes. Other than memory read/write, assume the processing delay is 0 ms on all nodes. The DLC layer adds an overhead of 125 Bytes to the PDU, and the physical layer adds no additional overhead. Assume there is no queueing on any link before the transmission. Answer all time-related questions in milliseconds (ms) . Note 1: Draw a time figure of the PDU(s) through all layers/links to determine the delays. Note 2: The DLC layer must completely read a packet from shared memory before transmitting it on the bitpipe, similarly the network layer software must completely read a packet from shared memory before forwarding it, and each layer must wait until a packet is completely available in shared memory before starting to read it. A. (3 pts) Assume there is no other transmission sharing any of the links. What is the bottleneck link bandwidth for this transmission? 100 Mbps. Bottleneck link is the one with the minimum bandwidth along the transmission path. B. (3 pts) At what time T1 does the PDU leave A? A PDU leaves a router when it has been fully transmitted on to the next link. Before leaving the node, the PDU incurs: 1) processing delay on A, and 2) transmission delay onto A-R1. Delay from NET to shared memory = 1000 Bytes * 8 / 1 Gbps = 0.008 ms (NET layer write to shared memory)

Your preview ends here

Eager to read complete document? Join bartleby learn and gain access to the full version