<title>Network Knowledge (transcription)</title> Network knowledge (transcription) Table of Contents
- Ethernet
- Carrier monitoring multi-point access/collision detection
- Token Ring
How data is exchanged
- Circuit switching: The bits of the entire message are continuously focused from the source, as if they were transported in a pipe
- Message exchange: The entire message is first transmitted to neighboring nodes, all of which are stored down to find the forwarding post and forwarded to the next node.
- Packet switching: A single packet (part of a message) is transmitted to adjacent nodes, stored down to find a forwarding post, and sent to the next node.
advantages of using packet switching :
Macro-Reliable: transmission process problems, you can go other paths (routers)
If the circuit is switched, one of the single points in the middle fails, resulting in an overall failure
Error handling can also be achieved
disadvantages of using packet switching :
Micro-unreliable: one of the packet transfers fails, then there is a problem. In addition, the possibility of loss of messages, chaos, repetition, etc.
Rate: Unit b/S, 1M b/s = 10^3k b/s, not stored in 2 binary way
Throughput throughput: The amount of data in a network per unit of time
Latency delay, latency: The time it takes to send to the receiving end of processing
Send delay the first bit of data sent, to receive the last bit to send the time required to transmit the delay propagation latency data to transmit the electromagnetic wave in the channel to propagate a certain distance the time processing delay The time that the receiver received the packet and processed the queue delay before entering the router to wait for processing in the input queue, the time required
TTL Time to Live: Live
RTT round-trip time: round trip times
Ethernet
Ethernet does not number the data frames sent, nor does it require the caller to send back a confirmation
This frame is discarded when the destination station receives a data frame with errors, and nothing else is done. The correction of errors is decided by the high level.
For example, if it is a TCP protocol, you will find that some data is lost and re-routed, and Ethernet does not know that this is a retransmission frame, but rather as a new data frame to send
Carrier monitoring multi-point access/collision detection
Carrier sense multiple Access with collision Detection CSMA/CD
As long as there is a computer on the bus to send data, the bus transport resources will be occupied, so at the same time can only allow one computer to send data, otherwise
Each computer interferes with each other, causing the data sent to be corrupted.
The site to which the data is transferred is first monitored for carrier-free media to determine if there are other sites that are transmitting data. If the media is idle, the site can transfer data
Otherwise, the site will be evasive for a period of time before trying. This requires a Backoff algorithm to determine the time of avoidance.
Commonly used Backoff algorithms have non-insistence, 1 insist, P adhere to three kinds of
Non-stick algorithm
- If the media is free, you can send it immediately.
- If the media is busy, wait for a random recurrence delay that is determined by the probability distribution, and then repeat the previous step
Using random recurrence delay can reduce the likelihood of conflict
The disadvantages of non-stick algorithms are:
Even though there are several points to be sent, the media may still be idle due to the delay in waiting, and the usage rate is reduced.
1-Persistence algorithm
- If the media is free, you can send it immediately
- If the media is busy, continue listening until the media is detected to be idle, send immediately
- If there is a conflict (not receiving a positive response for a period of time), wait for a random amount of time, repeat steps 1, 2
The advantage of this algorithm is: As long as the media idle, the site can be sent immediately, to avoid the loss of media utilization
The disadvantage is: if there are two or more than two sites have data to send, the conflict will inevitably
P-Persistence algorithm
- Listen to the bus, if the media is idle, then the probability of P is sent, and the probability of (1-P) delay a time unit
A time unit is usually equal to twice times the maximum propagation delay.
- After delaying a unit of time, repeat steps 1
- If the media is busy, continue listening until the media is idle and repeat step 1
The P-persistence algorithm is a compromise scheme that can reduce the media idle time as well as the 1-persistence algorithm as a non-stick algorithm.
The problem is how to choose a valid value for P, which takes into account the unstable state of the system in order to avoid heavy loads. If the media is busy, there are N stations with data waiting to be sent,
Once the current send is complete, the total expected number of stations that will attempt to transmit is NP. If you choose P too large to make np>1, indicating that there are multiple sites trying to send, the conflict
Inevitable. The worst case scenario is that as the probability of conflict increases, the throughput is reduced to zero
Therefore, the appropriate P-value must be chosen to make the np<1. Of course, the P value is too small, then the media utilization will be greatly reduced
Token Ring
The small data (frames) that are transmitted on the token ring are called tokens, and whoever has the token has the transfer permission. If a workstation on the ring receives a token and has information sent,
It changes one of the tokens (the action turns the token into a frame start sequence), adds the information that you want to transfer, and then sends the entire message to the next workstation in the ring
When this information frame is transmitted over the ring, there is no token in the network, which means that other workstations must wait for the data to be transferred. Therefore does not occur in the Token Ring network
Transport conflicts
Network Knowledge (transcription)