COMP 3825 Networking and Information Assurance Midterm Review Flashcards

Flashcards covering network structure, packet/circuit switching, QoS, layering, transport protocols (TCP/UDP), reliable data transfer, and congestion control mechanisms.

Network Edge

Part of the network structure consisting of hosts, such as clients and servers, which are often located in data centers.

Network Core

A mesh of interconnected routers that form a "network of networks" to facilitate packet forwarding.

Packet-switching

A network core technique where hosts break application-layer messages into packets and forward them across links, with each packet transmitted at full link capacity.

Store-and-forward

A packet-switching principle where the entire packet must arrive at a router before it can be transmitted on the next link.

One-hop Transmission Delay

The time calculated as $\frac{L}{R}$ to transmit an L-bit packet into a link with a transmission rate of R bps.

Circuit Switching

A network core alternative where end-to-end resources are reserved and dedicated for a specific "call," ensuring guaranteed performance with no sharing.

FDM (Frequency-Division Multiplexing)

A circuit switching technique where the link's frequency spectrum is divided among users.

TDM (Time-Division Multiplexing)

A circuit switching technique where users share a link by dividing time into slots.

Nodal Processing Delay ($d_{proc}$)

The time spent at a router checking for bit errors and determining the output link, typically less than a millisecond.

Queuing Delay ($d_{queue}$)

The time a packet waits at an output link for transmission, depending on the congestion level of the router.

Transmission Delay ($d_{trans}$)

The time needed to push all the packet's bits into the link, calculated as $\frac{L}{R}$ bits per second.

Propagation Delay ($d_{prop}$)

The time for a bit to travel from the start of the physical link to the next router, calculated as $\frac{d}{s}$ where $s \approx 2 \times 10^8\,m/s$.

Throughput

The rate in bits/time unit at which bits are transferred between a sender and receiver, categorized as either instantaneous or average.

Bottleneck Link

The link on an end-to-end path that constrains the total end-to-end throughput.

Physical Layer (Layer 1)

The layer responsible for encoding bits to send them over a single physical link via voltage levels, RF modulation, or photon intensities.

Link Layer (Layer 2)

The layer responsible for framing and transmission of collections of bits into individual messages sent across a single subnetwork.

(Inter)Network Layer (Layer 3)

The layer that bridges multiple subnets to provide global addressing and end-to-end internet connectivity (e.g., IPv4, IPv6).

Transport Layer (Layer 4)

The layer providing end-to-end communication between processes using protocols like TCP or UDP.

Application Layer (Layer 7)

The layer where communication of any freely structured data occurs, using protocols like HTTP, SMTP, or DNS.

IP (Internet Protocol) "Best Effort" Service

A delivery service where packets may be lost, corrupted, or delivered out of order.

Client-Server Architecture

An application structure with an always-on host (server) having a permanent IP address and clients that communicate with it but not directly with each other.

P2P (Peer-to-Peer) Architecture

An architecture where arbitrary end systems directly communicate and share service capacity without an always-on server.

Socket

The interface or "door" between an application process and the transport infrastructure controlled by the OS.

SMTP (Simple Mail Transfer Protocol)

An application-layer protocol [RFC 2821] used for email that operates over TCP.

HTTP (HyperText Transfer Protocol)

A web application protocol using request/response messages; GET is used to download data and POST is used to upload data.

DNS (Domain Name System)

A hierarchical distributed database that maps host names to numerical IP addresses.

Demultiplexing

The process at the receiver of using header info (IP addresses and port numbers) to deliver received segments to the correct socket.

TCP 4-tuple

The set of four values used to identify a TCP socket: source IP address, source port number, destination IP address, and destination port number.

UDP (User Datagram Protocol)

A "no-frills," connectionless transport protocol providing "best effort" service where segments may be lost or out of order.

UDP Checksum

A value used to detect bit-errors in a transmitted segment by treating the segment contents as a sequence of 16-bit integers.

Pipelining

A technique where a sender allows multiple "in-flight" yet-to-be-acknowledged packets to increase link utilization.

Go-Back-N (GBN)

A pipelined protocol where the sender can have up to N unacked packets and uses cumulative ACKs; a timeout triggers retransmission of all unacked packets.

Selective Repeat (SR)

A pipelined protocol where the sender maintains a timer for each unacked packet and retransmits only those for which an individual ACK was not received.

MSS (Maximum Segment Size)

The maximum amount of data that can be placed in a single TCP segment.

EstimatedRTT

A smoothed average of RTT measurements calculated as $(1 - \alpha) \times EstimatedRTT + \alpha \times SampleRTT$, with a typical $\alpha = 0.125$.

TCP Timeout Interval

The calculation defined as $EstimatedRTT + 4 \times DevRTT$ to set a safety margin for retransmission.

TCP Fast Retransmit

A mechanism that triggers retransmission of an unacked segment after the sender receives triple duplicate ACKs (3 ACKs for the same data).

TCP Flow Control

The process of a receiver advertising free buffer space ($rwnd$) to ensure the sender does not overwhelm the receiver's buffer.

TCP 3-way Handshake

The connection establishment process involving a SYN, followed by a SYNACK, and finally an ACK.

AIMD (Additive Increase Multiplicative Decrease)

The approach where TCP increases window size by 1 MSS every RTT until loss occurs, then cuts the window in half.

TCP Slow Start

A phase where the congestion window ($cwnd$) starts at 1 MSS and doubles every RTT until the first loss event or a threshold is reached.

TCP Tahoe

A congestion control version that always sets the congestion window to 1 MSS upon detecting a loss event (timeout or 3 duplicate ACKs).

TCP Reno

A congestion control version that sets $cwnd$ to 1 MSS on timeout but cuts it in half when 3 duplicate ACKs are received.

TCP CUBIC

A variation that ramps up to the sending rate just before the last congestion loss ($W_{max}$) quickly, then approaches it more slowly.

QUIC (Quick UDP Internet Connections)

A UDP-based protocol that supports multiple application-level streams and establishes reliability, security, and congestion control in a single RTT.