QUIZ 1 Assume fast re-transmission and fast recovery are enabled at the sender, and TCP is in congestion avoidance phase. Initially the cwnd equals 4, so packets 1-4 are sent out sequentially in a row. Packet 3 is dropped during transmission. 1.) What are the sequence numbers in the acknowledgement sent back from the receiver? Assume that when there is no packet loss, the receiver normally replies with Ack n+1 after receiving Pkt n. 2.) What happens next, assuming no further packet losses? Please show when and how packets and acknowledgements will be exchanged until Ack 7 is sent back from the receiver.

QUIZ 2 1.) Split horizon is a feature of? Insuring a common intra-domain protocol 2.) Which of the following is not true for Bellman-Ford algorithm? No hop count limit 3.) Which of the following is not a task of BGP? Distance vertor routing 4.) Which method does RIP use for determining routes? Restart 5.) Which of the following is not part of routing failure recovery? RIP QUIZ 3 1.) IP=OSPF re-convergence, SONET= automatic protection switching, WDM=fiber protection, MPLS=LSP protection 2.) Contrast protection vs. restoration in terms of a.) recovery approach, Protection allocates spare resource for backup, restoration computes network configuration on the fly b.)recovery speed, protection offers faster recovery than restoration c.)resource efficiency, restoration utilizes resources more efficiently 3.) What is the most significant difference between the 1:1 and 1+1 protection schemes, 1+1 establishes two active paths and signals are duplicated and transmitted on two paths simultaneously, while in 1:1 one path is used for backup only. 4.) Is the Internet a scale-free network or an exponential network? Are scale-free networks more prone to attacks than exponential networks? Why? Internet is a scale-free network. A scale-free network is more prone to targeted attacks because its connectivity is largely contributed by a few hot nodes with many links. QUIZ 4 1.) Given the following dimensions of a generic Portland data center network of size k, plot the network topology when k=4. You may start with the partial figure below. a.) number of ports per switch: k, b.) number of core switches: (k/2)^2, c.) number of pods: k, d.) number of edge switches per pod: k/2, e.) number of aggregate switches per pod: k/2, f.) total number of servers: k^3/4 2.) What are two main ideas of DCTCP that make it different from traditional TCP? A.) react in proportion to the extent of congestion. Reduce window size based on fraction

of marked packets. B.) Mark based on instantaneous queue length. Fast feedback to better deal with bursts. Simplifies hardware.

QUIZ 5 1.) Explain the differences between independent use and correlated use of multiple SDN controllers. Do they perform the same tasks? Do they need to communicate with each other? Give one example for each type. Multiple controllers can be used in two ways. Independent: different controllers perform different tasks, controllers do not need to communicate with each other, e.g. a controller for routing and another for address management. Correlated: controllers do need to communicate with each other, the major objectives include load balancing and resilience, e.g. the network is divided into subnetworks, each subnetwork has its own controller. 2.) Identify two advantages and two disadvantages of SDN. Advantages: No longer designing distributed control protocols, much easier to write/verify/maintain, NOS serves as fundamental control block with a global view of network. Bottlenecks: Bottleneck at the controller: communication channel between switches and the controller (i.e. control network). Bottleneck at the switch control plane: communication channel between switch forwarding plane and switch local CPU. The centralized controller is prone to targeted attach like DDoS. Lesson 1=Introduction DSL=Digital Subscriber Line o Link Bandwidth= twisted-wire remains the same, modem is different o POTS, ISDN, ADSL UpStream and DownS. VDSL=Very high bit-rate DSL o TCM-ISDN (reduces interference with ISDN line), VDSL 1st Step to High Speed: Link o Advanced technologies ! Broaden the road o Cost control + massive production ! Affordable equipment

With high-speed links o a fast computer is conn. To a slow server via good links o a group of computers with 100/10 M Eth connected to a switch

2ndStep: High-performance nodes o Terminals ! Directly affect the application performance o Network Nodes ! Determine the performance of data exchange

3rdStep: Powerful control and management o Fixed network resurce ! Maximize the utilizatio o Fixed performance criteria ! Minimize the investment and maintainance o Fixed network architecture ! Accelerate service provisioning ! Fast failure restoration o Stability of network

High Speed: Foundations

Application o Interconnection of different types of networks o Hierarchical architecture Links of Backbone Networks 1 o Time-domain multiplexing ! Signals are interleaved in the time axis ! Periodic slots turns to be a channel with fixed bandwidth ! A link can be divided into multiple channels

Links of Backbone Networks 2 o Wavelength-domain multiplexing ! A single fiber contains several wavelengths ! Each wavelength are relatively independent

Links of Backbone Networks 3

Backbone Networks o A fiber has a huge capacity, e.g. 320 Gbps o Wavelengths have coarse granularity, e.g. 10Gbps per wavelength o Applications need fine granularity, e.g. 1 Gbps o Question, How to handle the mismatch? Think about subway 1 o A fiber has a huge capacity, e.g. 320 Gbps Think about subway 2 o A tunnel has multiple trains o Each train can be regarded as a virtual tunnel Back to HighSpeedNetworks o Physical Topology ! Fibers ! OXC (optical cross connect) o Logical Topology ! Channels (slot channel or wavelength) ! Routers

Benefit of Logical Topology o Cost-effective o Better delay performance for cut-through logical links o Easy to manage packet streams

Switching Node

Packet switching

Lecture 2 Flow and Congestion Control Flow Control o Data transfer TX->RX, reliable link o Speed mismatch between TX and RX o How to realize speed adaptation? o Basic flow control: ! Stop-and-wait protocol ! Sliding window protocol

Stop-and-wait Flow control o Source transmits frame o Destination receives frame and replies with acknowledgement o Sourcewaits for ACK before sending next frame Sliding window flow control o Allow multiple frames to be in transit o Receiver has buffer W frames long o Transmitter can send up to W frames without ACK o Each frame is numbered o ACK includes number of next frame expected Transport Layer Flow Control o Performs end-to-end flow control across the network o Necessary for QoS