Networks Exam 1

why were networks first needed

terminals had very little computing power and needed to be networked/connected to the bigger machine for actual compute power. needed network/network adjacent architecture to support

encoding schemes (generally)

dictate how signals are represented through a physical medium. mostly deals with understanding mechanics.

concept of connection

networked devices need to have an understanding of senders and receivers.

standards for protocol implementation

using interfaces allows for a more malleable standard that can be implemented in many ways, and supporting good longevity. the downside is that implementations differing can cause confusion while still technically satisfying the interface.

what is the whole idea behind networks

provide a layered set of protocols / a laered set of peer to peer interfaces

OSI model layers

application, presentation, session, transport, network, data link, physical

OSI application layer

An application. Hard to define exactly because the point of application being at the top is that it can be whatever it wants to be.

OSI presentation layer

Dictates data formatting.

OSI session layer

manages/organizes communication sessions between applications

OSI transport layer

process to process delivery. how to handle bytes as they are delivered. help ensure reliability.

OSI network layer

network/global address system to identify who to communicate with.

OSI data link layer

Intermediary needs to identify devices through ports (physical MAC addresses often). Needs to map network and physical addresses.

OSI physical layer

needs to communicate with other devices over some sort of hardware. this is the hardware connection between devices.

Internet Architecture Layers

application, transport, internet, physical

Internet Architecture Application layer

An application. Hard to define exactly because the point of application being at the top is that it can be whatever it wants to be. also includes the protocols dictating the data formatting as well as managing/organizing communication sessions between applications

Internet Architecture Transport Layer

process to process delivery. how to handle bytes as they are delivered. help ensure reliability.

Internet Architecture Internet Layer

network/global address system to identify who to communicate with.

Internet Architecture Physical layer

Intermediary needs to identify devices through ports (physical MAC addresses often). Needs to map network and physical addresses. needs to communicate with other devices over some sort of hardware. this is the hardware connection between devices.

headers

wrap frames in information for the current layer as we down the architecture. moving back up, the headers are removed and read.

footers

may be included in a message as an integrity check

payload

the actual data or contents being sent over a network in a packet.

what is the payload for a request?

requests often have no plload and instead the headers do all that is needed to indicate that the message is a request.

common application protocols

kind of a trick question cause the are so varied, but like maybe HTTP for web pages.

common transport protocols

TCP, UDP

common network protocols

IP is basically the only one, particularly for IP architectures cause duh.

common physical protocols

like ethernet and wifi and stuff

ARP

a way of forcing a mapping of physical devices to Network devices. more specifically, it maps IP/network-layer addresses to MAC/link-layer addresses on the local network.

are all four layers of the internet architecture mandatory?

no. goes bottom to top. can’t have higher layers with the layer below, but that doesn’t mean you need to have the higher layers.

bandwidth (network performance)

measures the number of bits sent a second. technically the maximum/rated data transmission and often doesn’t actually reach this number.

latency

the amount of time it takes a signal to propagate over a physical medium, queue at different points in the network, and be otherwise processed.

propagation delay

the time it takes a physical signal to transmit over a medium.

adapter

a mechanism that allows analog signals to be transmitted. Examples include ethernet ports and wifi cards. plugs into the host, owns the MAC address ,builds and reads frames.

digital signals vs analog signals

digital signals exhibit discrete peaks and valleys whereas analog signals are more continuous and sinusoidal.

Non-return to Zero

Ones are represented by high voltages and zeros by low voltages. for a string of 1s, the signal remains high instead of dropping back low. The pro of this is that it is simple and requires relatively low power consumption. The lack of changes in voltage can, however, leave the encoding scheme susceptible to baseline wander and make synchronizing clocks more difficult .

clock

“a metronome for bits.” a beat at which data is conveyed.

why must clocks be synchronized

if one computer is sending at its clock rate and another is reading at its different clock rate, bits are likely to be misinterpreted.

converting to analog signals

computer signals/bits come in discrete values but physical mediums are not nearly so discrete. as a result, typically values above a certain threshold are considered “high” and below are considered “low” to have the analog signal function in a way more like a digital signal. usually, it is voltage modulation, but sometimes it is wavelength modulation or phase modulation.

baseline wander

the errors that occur with the moving average. After a series of consecutive high signals, the baseline may move too high and some intended 1s may be interpreted as 0s. After a series of consecutive low signals, the reverse may happen.

moving average

a sliding window determines the most recent voltage levels and takes the average to be the threshold that the signal must be over or under to be considered high vs. low.

phase locked loops

on the adapter, an electronic feedback control system that references the phase of an incoming signal to find a common speed in between systems and translate to the common clock speed and phase. phase

point of encoding scheme is to

solve the two issues of voltage variance/baseline wander and asynchronicity.

can encoding schemes differ within the same protocol

yes. ethernet devices, for example, may use any of the encoding schemes.

Non-return to zero inverted

encoding scheme where staying at the same voltage level indicates a 0, whereas a change in voltage level indicates a one.

what is the one thing needed to solve the main two problems in an encoding scheme

more bit transitions

Manchester encoding

an encoding scheme where 0 and 1 are represented by low to high transitions and high to low transitions (or the reverse according the IEEE standards for manchester inverted.) Has the maximum number of transitions cause everything needs to be a transition. takes twice as long to transmit anything though (sending two 0s in a row requires that it go low to high twice so at some point it has to go high to low to “reset”). So at a cost of less bit efficiency, it is really good for issues with baseline wander and synchronization.

4B/5B

encoding scheme based on non-return to zero inverted (NRZI). Each 4 bit sequence is translated to a predetermined 5 bit sequence. Since there are more 5 bit sequences than 4 bit sequences, the 5 bit sequences with good transitions can be selected and used. not the most number of transitions or 100% bit efficiency, but it is a really good trade off with 80% bit efficiency and typically enough transitions to get around baseline wander and synchronization issues.

frame

blocks of data on packet switched networks.

how are frames nested

the body of one frame contains the frame of the next layer. For example, a phy

what is a sentinel character

a character used to signify the beginning/end of a frame. ~ is one that is sometimes used.

what is framing

one implementation of packet format; defines how the beginning/end of a block of data is identified.

what is a byte-oriented protocol?

a framing protocol where special bytes/characters mark frame boundaries. P2P is an example.

what problem do sentinel characters cause?

the sentinel might occur naturally in the body, making it look like a frame boundary.

how do byte-oriented protocols solve sentinel characters appearing in the body?

escape sequences: an escape character indicates the following sentinel is data, not a boundary.

what is a bit-oriented protocol?

frames are identified using a specific bit pattern rather than special character/bytes.

How does bit-oriented framing prevent its boundary pattern from appearing in the body?

bit stuffing: extra bits are inserted into the body so the boundary pattern cannot occur accidentally.

what is the advantage of bit-oriented framing?

the frame format does not depend on character/byte boundaries.

What is CRC used for in a frame?

cyclic redundancy check. dtects whether data changed or was corrupted during transmission. typically 32 bits.

error detection

indicates an error has occurred. Does not say what error. All we know is someething has happened and we need to drop the whole frame.

IP Checksum

breaks the data into fixed-size blocks, often 16-bits. add the blocks with ones complement addition (overflow wraps around and gets added back in). invert all the bits of the result. put checksum in message. receiver does calculation again, and if the result is all ones it is good.

which is safer checksum or CRC

CRC is more likely to detect errors

which is less computationally intensive, checksum or CRC?

checksum

how is crc checked and generated

divide/xor using the agreed divisor. remainder of 0 means the frame passes. alternatively, to calculate it you find the remainder and put that in.

Transceiver

taps the cable, senses idle/busy (carrier sense), drives and reads the raw signal

difference between transceiver and adapter

the transceiver talks to the physical medium while the adapter connects the computer to the network. transceiver is sending and receiving the signal, and the adapter is the larger network interface that does things like error detection.

C(x)

the CRC divisor or generator. known by the sender and receiver. good ones can detect single bit errors, double bit errors, and odd numbers of errors (given certain conditions).

which IEEE standard is ethernet

802.3

carrier sense

every node can tell an idle link from a busy one before it tries to talk

multiple access

many nodes share one link, like several stations plugged into one bus

collision detection

a node listens while it sends, so it can notice its own frame getting stepped on.

repeaters

join segments into one larger shared network.

5-4-3 rule

5 segments, 4 repeaters, 3 mixing segments.

frame format stuff

preamble - 8 bytes, dest addr - 6 bytes, src addr - 6 bytes, type - 2 bytes, data - 46 to 1500 bytes (padded if below 46), CRC - 4 bytes. (64 byte total but for some reason preamble is not counted).

preamble

alternating 0s and 1s, lets the receiver sync to the signal

addresses

48 bits each , identify source and destination

type

the demultiplexing key, which higher-level protocol gets this frame

MAC address

on the adapter, not the host. written as six hex byte-pairs, separated by colons. leading zeros are dropped in the human-readable form.

unicast

one sender, one specific receiver. frame contains the MAC address of the intended adapter.

broadcast

one sender, everone on the local network. uses the broadcast MAC address (all ones); every adapter accepts it.

multicast

one sender, a specific group of receivers. adapters can subscribe/opt in to particular multicast addresses.

promiscuous mode

adapter accepts every frame it sees, even if it addressed to another device.

super basic transmission rule

line idle and a frame is ready, then transmit immediately, no negotiation. line busy, wait for idle then transmit. all adapters wait the transmission time of 96 bits before the next frame starts.

two nodes talk at once

carrier sense does not guarantee no collisions, propagation delay leaves a gap. what a colliding adapter does → jamming sequence then stop.

is collision a design flaw?

not really it’s pretty expected.

what does a host do upon detecting collision?

immediately sends a 32 bit jamming sequence. stops transmitting early. runt frame - 64 bit preamble and 32 bit jam.

Why can two hosts transmit even after sensing an idle medium

propagation delay - one host’s signal may not have reached the other yet

why does ethernet have a minimum frame size? what is this size.

sender must still be transmitting when the jamming signal reaches it to actually detect the collision, so the frame must be long enough to make this the case even when the nodes are the furthest away as possible. minimum size is 64 bytes.

round trip time

time for the signal to reach the farthest host and for the effect/response to return.

why was 64 bytes sufficient for 10 Mbps ethernet?

512 bits takes 51.2 micro seconds to send, long enough to cover the worst-case RTT.

near vs. far collision

near - detected quickly. far - collision may take almost an RTT to return to sender.

p-persistent protocols

not really used (there but configured to be off). timeis divided into slots, one frame length each. idle slot - transmit with probability p, defer with probability 1-p. if it defers and the next slot is idle it “rolls the die” again to see if it can go.

exponential backoff

when there is a collision, the frame picks how long to wait before trying to send again. typically between 0 and 2^n - 1 where n is the number of collisions. gives up after 16 collisions. collide, jam, stop, backoff, retry.

wifi as a medium

medium is radio waves. waves are broadcast so that anyone in range can hear. bit errors are even more common due to varying transmission power.

joining a wifi network

send a probe frame → broadcast connection request. response → access point replies with probe response. node sends association request to the strongest signal access point. the access point confirms with association response. handshakes are probably done - potentially to set up diffeehellman idk how to spell whatever.

active scanning

node sends a probe frame to hunt for an access point.

passive scanning

access points periodically send a beacon frame. nodes listen then associate.

wifi frames.

control - 2 bytes, address 1 - 6 bytes, address 2 - 6 bytes, address 3 - 6 bytes, address 4 - 6 bytes, data - 0 to 2312 bytes, CRC 4 bytes.

hidden node problem

if two nodes can reach a third node but not each other, signals can collide at the third node and A and C are unable to tell. B is also unable to let A or C know a collision occurred as the data is garbled.

exposed node problem

A-B-C-D are all nodes in a line. B is transmitting to A. C thinks the medium is busy when it wants to transmit to D, but really since C is outside A’s range and B is outside D’s range it would be fine. so basically, a node hears a nearby transmission and unnecessarily decides it can’t transmit.

CSMA/CD vs CSMA/CA

the cd is collision detection. it is what ethernet tends to do. the ca is collision avoidance and is what wifi does since it can’t reliably detect collisions while transmitting.

RTS/CTS

collision avoidance technique used by wifi. ready to send frame is sent by sender. if the receiver is available, it replies with a clear to send. all neighbors hear this too. the sender then transmits the data and the receiver acks to confirm it got it. the double confirmation of RTS/CTS keeps the network clearer. Since the RTS frame is shorter, there is less likely to be overlap. CTS will also specify who is clear to send.