For the analysis of the impact on the communication MTU, make the following tests:

Sends a packet size is 4000 byte, is 10000 packages, respectively under the MTU=1024 and MTU=4096 test,



The measured result is:

The bandwidth of the MTU=4096 is 105% of the MTU=1024

MTU=4096, when the time delay (latency) is 127% of the MTU=1024

Confused:

1, the same package size, so small delay should not be more bandwidth?

2, we know that when the MTU=1024, 4000 - byte packet will produce the subcontract, divided into three packages, and MTU=4024, will not produce the subcontract,

So the time delay, refers to the total package sent after the completion of the three subcontract to each other (i.e.,) response delay, or segmentation of the package, the first package to each other after each other reply delay?

3, get find information that is parcel data, small MTU of time delay, bandwidth is small, big delay of MTU, large bandwidth, how to understand this?

CodePudding user response:

1. The problem of bandwidth, data transmission must have agreement, agreement will be in baotou data, such as the actual transmission information 100 bytes, have a fixed agreement first 20 bytes, the efficiency is 100/120, 200 is 200/220, if transmission efficiency up this way, so the actual available bandwidth is big

2. Delay this should be well understood, all the data transmission time, general network data are serial data, according to the bit transmission, content of course big, the time is much, big delay

CodePudding user response:

reference 1st floor Yofoo response:

1.

Thank you for your reply, but you seem to understand the topic of my wrong, I'm asking is

Sends a packet size is 4000 byte, is 10000 packages, respectively under the MTU=1024 and MTU=4096 test,

MTU bandwidth than MTU=1024=4096, but the MTU=4096 delay actually also MTU=1024,


The same package size, so small delay should not be more bandwidth? In other words the same massage size, bandwidth should not be greater than the bandwidth of small time delay is lower? Internal energy is transferred in a 1 second delay small code more packets, more bandwidth,

CodePudding user response:

Size and bandwidth delay didn't any relationship unless the network load limit, causes the retransmission delay increase problems, you can't put this thing to other top
Such as satellite communications, the bandwidth can be done is very high, but the delay is very big

CodePudding user response:

reference Yofoo reply: 3/f

size and bandwidth delay didn't any relationship, unless to the upper limit of network load, will cause the retransmission delay increase problems, you can't put this thing to other top
Such as satellite communications, the bandwidth can be done is very high, but the delay is very large

A: 10 KB package delay 10 ms, so 1 second can pass 100 packets, bandwidth is 1000 KB/s

B: 1 MB of packet delay 20 ms, so 1 second can pass 50 package, also is the bandwidth of 50 m/s

Actually bandwidth and time delay has nothing to do at this time, but my situation is A, B package size is the same, that is,

A: 1 m package delay 10 ms, so A second can pass 100 packages

B: 1 MB of packet delay 20 ms, so 1 second can pass 50 bags

So this time, not delay the bandwidth of the small big? In turn the bandwidth delay is not small? Is that why my test result?