Internet Control Message Protocol creates messages for failu…

Questions

Internet Cоntrоl Messаge Prоtocol creаtes messаges for failure at which of the following layers?

Prоblem 1 (2 pоints) Mentiоn two key rаtionаle behind In-network computing. Problem 2 (3 points) Recаll implementing a key-value cache using Match-Action stages of a programmable switch. Suppose we have four entries in the cache – {Key: A, Value: A0 A1 A2}, {Key: B, Value: B0 B1}, {Key: C, Value: C0}, {Key: D, Value: D0 D1}. Below is the (partial) implementation of the cache using four Match-Action stages. The operations for reading and updating the cache are implemented using the process_array action shown below. Fill the empty bitmap and index fields in Stage 1 and the empty register array values in Stages 2-4. Recall a bitmap indicates the register arrays that store the key's value (i.e., a value of 1 for bitmap[i] indicates that register array i stores the key's value) and index indicates the slots in the array to get the value. Assume most significant bit (MSB) of bitmap is indexed 0. Problem 3 (3 points) Figure below shows the VMs for three tenants - tenant R (whose VMs R1, R2, R3 are shown in Red), tenant G (whose VMs G1, G2, G3 are shown in Green), and tenant B (whose VMs B1, B2 are shown in Blue). The minimum bandwidth guarantee as part of the Service Level Objective (SLO) for tenant R is 2 Gbps, tenant G is 5 Gbps, and tenant B is 3 Gbps. Each host has bandwidth capacity of 10 Gbps. There are four active flows in the network: R1(rightarrow)R3, R2(rightarrow)R3, B1(rightarrow)B2, and G1(rightarrow)G3. The rate for flow G1(rightarrow)G3 is 5 Gbps, while the rates for the remaining three flows are not known to you. Based on this information, please answer the questions below. (a) What will be the rate feedback sent by R3 and B2 to R1, R2, and B1 in EyeQ? Rate feedback from R3 to R1: _____ Gbps Rate feedback from R3 to R2: _ Gbps Rate feedback from B2 to B1: _ Gbps (b) What will the rates for flows R1(rightarrow)R3, R2(rightarrow)R3, and B1(rightarrow)B2 finally converge to in EyeQ? Rate for flow R1(rightarrow)R3: _ Gbps Rate for flow R2(rightarrow)R3: _ Gbps Rate for flow B1(rightarrow)B2: _ Gbps Problem 4 (2 points) Why does an RDMA network need Priority Flow Control (PFC)? Why do we need a congestion control protocol like DCQCN on top of PFC? Problem 5 (2 points) DCQCN is a congestion control protocol designed for lossless networks. But given that there are no per-packet ACKs in a lossless network, how is the sending rate of a flow increased in DCQCN? Provide both the condition for rate increase and the corresponding equation(s). Problem 6 (2 points) Mention two advantages of resource disaggregation inside datacenters. Problem 7 (2 points) Suppose you are a cloud provider and you receive a request to provision a VM with 10 GB of memory. Suppose you know that at any time, the VM only touches up to 6 GB of memory. Answer the questions below assuming you are using Pond's memory allocation algorithm. (a) Suppose the VM is running latency sensitive workload. How much memory should you allocate to the local node on which the VM is running and how much memory should you allocate to the remote memory pool? (1 point) Local memory: _ GB Remote memory: _ GB (b) Now answer the above question assuming that the VM is running latency insensitive workload. (1 point) Local memory: _ GB Remote memory: _____ GB Problem 8 (4 points) Consider a RotorNet network with 5 nodes – A, B, C, D, and E. Below is the static permutation schedule used to reconfigure the network. (a) Suppose nodes A, B, C, and D all send 1 packet each destined to final destination E at time 0. Fill the table below with the intermediate node that each packet goes through on its way to E. If a packet does not go through an intermediate node (i.e., it is delivered directly to node E), fill the corresponding table entry with "None". (2 points) (b) Following from the above question, at what time does each packet arrive at node E? Assume zero propagation delay, i.e., a packet transmitted at time (t) will reach the next hop at time (t) itself, and a node can only transmit one packet in each time slot. (2 points) Congratulations, you are almost done with this Final Exam. DO NOT end the Honorlock session until you have submitted your work. When you have answered all questions: Use your smartphone to scan your answer sheet and save the scan as a PDF. Make sure your scan is clear and legible. Submit your PDF as follows: Email your PDF to yourself or save it to the cloud (Google Drive, etc.). Click this link to submit your work: Final Exam Return to this window and click the button below to agree to the honor statement. Click Submit Quiz to end the exam. End the Honorlock session.