Basics of Networking Concepts

What’s a Network: Key Features

• Offering services, like transportation of goods, mail, information or data
• Sharing resources, used concurrently by multiple users
• Basic building blocks: nodes (active entities) that process and transfer data, and links (passive medium) as carriers
• Typically multi-hop, where two endpoints cannot directly reach each other, requiring other nodes to relay data

What’s a Network: “Nuts and Bolts” View

• Network edge comprises millions of end-system devices like PCs, workstations, servers, PDAs, phones, etc.
• Network core consists of routers and switches, forwarding data in packets (packet switching) or calls (circuit switching).
• Network edge devices run network applications and connect to network core through communication links like fiber, copper, or radio.

Switched Networks

• A network can be recursively defined as either two or more nodes connected by a link, or two or more networks connected by a node.

Brief History of the Internet

• Started in the 70s as a research project with 56 kbps speed and less than 100 computers.
• In the 80s, ARPANET and MILNET split, with NSF building NSFNET as the backbone, linking 6 supercomputer centers with 1.5 Mbps speed and 10,000 computers.
• From 87 to 90, regional networks linked with 100,000 computers.
• In the 90s, NSFNET moved to 45 Mbps and included 16 mid-level networks.
• In 1994, the NSF backbone was dismantled, leading to multiple private backbones.
• Today, backbones operate at over 10 Gbps speed and connect over 600 million computers across more than 190 countries.

Strategies

• Circuit switching: transmits bit streams, used in the original telephone network.
• Packet switching: stores and forwards messages, used in the Internet.

Addressing, Routing, and Forwarding

• Address: a byte-string used to uniquely identify a node.
• Types of addresses:
  • Unicast: specific to a single node.
  • Broadcast: addressed to all nodes on the network.
  • Multicast: addressed to a subset of nodes on the network
• Routing: the process of determining the path a packet will take through a network to reach its destination.
• Forwarding: the operation of sending an incoming packet to the correct switch or router output.

Multiplexing

• Synchronous Time-Division Multiplexing (STDM)
• Frequency-Division Multiplexing (FDM), e.g., TV channels

Statistical Multiplexing

• Used in the Internet.
• On-demand time-division, scheduling the link on a per-packet basis.
• Packets from different sources are interleaved on the link.
• Buffer packets that are contending for the link.
• Buffer (queue) overflow leads to congestion.

Inter-Process Communication

• Turns host-to-host connectivity into process-to-process communication.
• Bridges the gap between what applications expect and the underlying technology provides.

IPC Abstractions

• Request/Reply: used in distributed file systems, digital libraries (web).
• Stream-Based: transmits data as a sequence of frames, suitable for video applications like on-demand video and video conferencing.

What Goes Wrong in the Network?

• Bit-level errors: caused by electrical interference.
• Packet-level errors: due to congestion.
• Link and node failures can occur.
• Packets can be delayed, delivered out of order, or duplicated.

Layering Architecture

• Uses abstractions to hide complexity.
• Abstractions lead to layered architectures.
• Alternative abstractions exist at each layer.

Protocols

• Building blocks of network architectures.
• Define format, order of messages sent and received, and actions taken during message transmission and receipt.
• Each protocol object has two interfaces:
  • Service interface: operations on this protocol (e.g., how to obtain a message).
  • Peer-to-peer interface: messages exchanged with peer (e.g., message format definition.
• "Protocol" can refer to the specification of the peer-to-peer interface or the module that implements this interface.

Interfaces

• Interfaces between protocols can include a service interface for interactions within a host and a peer-to-peer interface for interactions with its protocol counterpart in another host

TCP/IP protocol suite

• A layered architecture with the following layers:
  • Application: supporting network applications like FTP, SMTP, HTTP.
  • Transport: enables process-process data transfer using protocols like TCP and UDP.
  • Internet: responsible for host-host data transfer through protocols like IP and routing protocols.
  • Data link: manages data transfer between neighboring network elements using protocols like PPP and Ethernet.
  • Hardware Devices (Physical): responsible for transmitting bits over the physical medium using the physical layer.

ISO/OSI model vs. TCP/IP

• Both models are layered architectures but have differences in their layer structure and naming conventions
  • ISO/OSI: Application, Presentation, Session, Transport, Network, Data Link, Physical
  • TCP/IP: Application, Transport, Internet, Data Link, Physical
• Certain layers in TCP/IP may map to multiple layers in ISO/OSI

Communication through Internet

• Data is transferred through multiple switches and routers in a network.
• Each device processes and forwards packets based on the IP address destination.

Identical objects between layers in TCP/IP

• Data is encapsulated from one layer to another, but the encapsulated data maintains the same object type across layers.
• e.g., segments or user datagrams in the Transport layer become datagrams in the Internet layer.

Encapsulation (Header/Body)

• Data is wrapped with a header in each layer containing information for that layer.
• The body contains the actual data that will be transmitted across networks.

Internet Architecture

• Defined by the Internet Engineering Task Force (IETF).
• Hourglass design, with application layer protocols (FTP, HTTP, etc.) and transport layer protocols (TCP, UDP) at the top, and the Internet layer (IP) at the bottom.
• Multiple networks connect to the internet through IP.

Implications of Hourglass

• A single Internet layer module (IP protocol) allows all networks to interoperate and all applications to function on all networks, regardless of underlying network technologies

Network Performance

• Latency: total time to transmit a data block.
• Factors affecting latency:
  • Initial time to establish a connection (setup time).
  • Time required to send the data block (transmit time).
  • Length of links and link speed (propagation delay).
  • Processing delay at routers (queuing delay).
• Overall network performance:
  • Throughput = Size-Of-Data-Block / Latency
  • Effective Bandwidth

Performance Metrics

• Transmit time:
  • Determined by data block size and link bandwidth.
  • Bandwidth: data transmitted per time unit (e.g., 10 Mbps).
• Propagation delay:
  • Determined by link length and link speed.
  • Propagation Delay = Distance / c (speed of signal).

Network Performance Revisited

• Total latency is calculated by summing up the setup time, propagation delay, transmit times, switch processing time, and RTT.
• Total latency is affected by factors such as handshake time, switch delays, and packet sizes. ### Network Performance
• Latency is the time it takes for a message to travel from source to destination.
• It consists of setup time, propagation time, transmission time, and switch processing time.
• Throughput is data transferred over time.
• Delay x Bandwidth product is the amount of data in flight or in the pipe, typically relative to Round Trip Time (RTT).
• For a 1-byte message, latency dominates bandwidth.
• For a 25MB message, bandwidth dominates latency.
• Infinite bandwidth makes RTT the dominant factor for throughput.
• Effective Bandwidth is the total transfer size divided by total transfer time.
• Transfer time consists of RTT and transfer size divided by bandwidth.

Other Metrics

• Variable Bit Rate (VBR) and Constant Bit Rate (CBR) are bandwidth requirements for applications.
• Delay jitter can be mitigated with buffering.
• Reliability is measured by packet loss rate.

Socket API

• The Socket API is a set of system calls for network communication.
• It implements network communication between processes.
• Socket creation uses the socket() function.
• Domains include PF_INET for the Internet Protocol (IP) and PF_UNIX for local inter-process communication.
• Types include SOCK_STREAM for connection-oriented communication, SOCK_DGRAM for connectionless communication, and SOCK_RAW for low-level access.
• connect() is used for active opens, initiating a connection from a client.
• bind() is used for passive opens, configuring a server to listen for connections.
• listen() specifies the maximum number of pending connections.
• accept() establishes a connection from a server to a client.
• send() and recv() are functions for sending and receiving messages, respectively.

Example Socket Programs

• Client side programs use socket, connect, send, and close.
• Server side programs use socket, bind, listen, accept, recv, and close.

Network Features

• Networks provide services to transfer goods, mail, information or data.
• Networks are shared resources, often used concurrently by many users.
• Networks are built from nodes (active entities) that process and transfer data in addition to links (passive medium) that carry the data.
• Often networks are "multi-hop," meaning end points cannot directly reach each other and intermediate nodes are required to relay data.

Network Architecture

• Network edge is composed of end system devices, like PCs, workstations, servers, PDAs, phones, and toasters.
• Network core is composed of routers and switches, which forward data in packets or circuit-switched calls.
• Communication links include fiber, copper, radio, and satellite.
• Links can be point-to-point or have multiple access.
• A network is defined recursively as two or more nodes connected by a link or two or more networks connected by a node.

History of the Internet

• The internet was created in the 70's as a research project with a limited number of computers.
• The ARPANET and MILNET split in the 80's.
• NSFNET was built as a backbone in the 80's and connected to other networks and institutions and included supercomputer centers.
• NSFNet moved to a higher bandwidth in the 90’s.
• The NSF backbone was dismantled in the 90’s and replaced with private backbones.
• Today, internet backbones have high bandwidth and connect to a large number of devices worldwide.

Strategies

• Circuit switching: Data is transmitted continuously along a dedicated path.
• Packet switching: Data is divided into packets, and these packets are forwarded independently and routed through various paths in the network.

Addressing

• Each node on the network has a unique identifier called an address.
• Types of addresses:
  • Unicast: Unique to one node.
  • Broadcast: Sent to all nodes on the network.
  • Multicast: Sent to a specific subset of nodes on the network.

Routing

• Determines the path packets will take through the network to reach their destination.

Forwarding

• Operations to send an incoming packet to the correct output on a switch or router.

Multiplexing

• Allows multiple users to share the same physical transmission medium.
• Synchronous Time Division Multiplexing (STDM) and Frequency Division Multiplexing (FDM) are examples.

Statistical Multiplexing

• Used in the internet.
• It’s on-demand time-division, meaning the link is scheduled on a per-packet basis, with packets from different sources interleaved on the link.
• If a packet arrives at the link when it's already occupied, it's buffered in a queue.
• When buffers overflow, congestion occurs.

Inter-Process Communication

• Turns host-to-host connectivity into process-to-process communication.
• Allows applications on different hosts to communicate with one another.
• Fill gaps between applications and the underlying technology.

IPC Abstractions

• Request/Reply: A common communication pattern for request-response interaction, like distributed file systems or web servers.
• Stream-Based: Used for streaming applications like video conferencing, where information is continuously transferred as a sequence of frames.

Network Errors

• Bit-level errors caused by electrical interference.
• Packet-level errors caused by congestion, leading to delays, out-of-order delivery, and duplicates.

Layered Architecture

• Uses abstractions to hide complexity and simplifies network management.
• Network operations are divided into layers, with each layer responsible for a specific aspect of networking.

Protocols

• Define the format, order of messages, and actions taken during transmission and receipt between network entities.
• Each protocol object has a service interface (operations on that protocol) and a peer-to-peer interface (defining the messages exchanged between protocol peers).

TCP/IP Suite

• A collection of networking protocols used on the internet and other networks.
• It consists of four layers:
  • Application Layer: Provides services to support network applications (FTP, SMTP, HTTP).
  • Transport Layer: Enables process-to-process data transfer (TCP, UDP).
  • Network Layer: Manages host-to-host data transfer (IP, Routing protocols).
  • Data Link Layer: Provides reliable data transfer between neighboring network elements (PPP, Ethernet).
  • Physical Layer: Transmits bits across the physical medium.

ISO/OSI Model

• A layered model that describes the interaction of communication protocols.
• It is very similar to the TCP/IP suite.

Internet Architecture

• Consists of networks and hosts connected using layered communication protocols.
• Defined by the Internet Engineering Task Force (IETF).
• Hourglass Design is the common architectural model.
• The top layer is Application Layer and allows for various application protocols (FTP, HTTP, NV, TFTP).
• The middle layer is Transport Layer and provides services like TCP and UDP.
• The bottom layer is Internet Layer which defines the IP protocol for host-to-host data transfer.

Implications of Hourglass Design

• A single Internet layer module allows all networks that support IP to interoperate.
• Allows all applications to function on all networks that support IP.
• Enables simultaneous development of network applications and underlying technologies.

Network Performance

• Latency measures the time it takes to transfer data.
• Factors influencing latency include: Setup time to establish communication, transmit time to send data, propagation delay across links, and queuing delay in routers.

Delay Measurements

• Transmit time: Time to put data on a link, dependent on the link's bandwidth.
• Propagation delay: Time for signal to travel across the link and depends on link length and signal speed.

Network Performance Revisited

• To transfer a 1.5 MB file from host A to Host B through a switch S1, the total time is calculated as the sum of the setup time, propagation delay, transmit time on both links (A to S1 and S1 to B ), and switch processing time.
• This calculation also includes the handshake time required to establish communications (2 x Round Trip Time).

Latency Calculation

• Latency is the time it takes for a message to travel from one point to another.
• It includes setup time, propagation time, transmission time, switch processing time.
• In the example (A to S1 to B) the latency is: 0.16s + 0.04s + 1.258s + 0.8192 ms + 0.01s = 1.469s

Throughput Calculation

• Throughput is the amount of data that can be transferred per unit of time.
• In the example the file size is 8 x 1.5 x 220 bytes = 2,640 bytes
• The throughput is: 8,565,631 bps = 8.566 Mbps

Delay x Bandwidth Product

• The delay x bandwidth product represents the amount of data "in flight" or "in the pipe"
• It is usually relative to the Round Trip Time (RTT)
• Example: 100ms x 45 Mbps ≈ 560 KB

Bandwidth vs Latency

• The relative importance of Bandwidth and Latency depends on the size of the message
• For 1 byte messages, Latency is more important than Bandwidth
• For large messages (e.g., 25 MB), Bandwidth is more important than Latency.
• Infinite bandwidth: RTT dictates the throughput.

Other Metrics

• Bandwidth need: VBR (variable bit rate) vs CBR (constant bit rate)
• Delay jitter: requires buffering
• Reliability: packet loss rate
• Efficiency: Overhead of implementation

Socket API

• Socket API provides functions for network communication
• socket() function creates a socket.
• connect() function initiates a connection (active open).
• bind() function associates a socket with an address (passive open).
• listen() function puts the socket in a listening state (passive open).
• accept() function accepts an incoming connection (passive open).
• send() function sends data to a socket.
• recv() function receives data from a socket.

Example Socket Program (Client)

• Code snippet demonstrates a simple client-side socket program.
• sin.sin_addr.s_addr is the server's IP address.
• sin.sin_port is the server's port.
• The code creates a socket, connects to the server, sends data, and then closes the socket.

Example Socket Program (Server)

• Code snippet demonstrates a simple server-side socket program.
• The server binds to a specific address and port, listens for incoming connections, accepts the connection, receives data, processes the data, and then closes the socket.

Network Software Introduction

• Covers applications, features, statistical multiplexing, inter-process communication, network architecture, performance metrics, and implementation issues.