It can take a while depending on the size of the document..please wait
Discuto
Starting: 22 Aug
Ending
0 days left (ends 29 Aug)
Go to discussion, participate and give your opinion
description
Further info
LATEST ACTIVITY
LEVEL OF AGREEMENT
MOST DISCUSSED PARAGRAPHS
No activity yet
LATEST COMMENTS
No activity yet
MOST ACTIVE USERS
No activity yet
Status:
Closed
Privacy:
Public
CONTRIBUTORS (2)
Share:
_
filter:
All Most voted (0) Most comments (0) Controversial (0) Decided (0)
In voting (0)
Supported (0) My contributions (0)
view:
<< Previous paragraphs
P70
• In a web browser environment it is difficult or even impossible to setup UDP-based connections.
Add comment
P74
• calculated as an average of recorded round-trip time values (typically expressed in milliseconds).
Add comment
P75
The measurement setup should be insensitive to (user) clock changes during the measurement.
Add comment
P76
It is also recommended that the delay variation (jitter) is calculated as mean deviation based on the samples collected for the delay measurement.
Add comment
P77
Calculation shall be based on the algorithms used in the Linux ping utility which is based on 4.3BSD
Add comment
P82
If a packet is not received back within a certain timeout (e.g. 3 seconds), it is considered as lost for the purpose of packet loss measurements.
Add comment
P83
Due to the effect of packet loss on TCP connections, even the low level of packet loss observed in modern networks[2] can result in significant performance degradation. Therefore it is evident that 10 or even 100 measurements may not yield packet loss accurately. It is therefore recommended to send a large number of IP packets (e.g. at least 1000). The number of IP packets should be based on access technology characteristics.
Add comment
P84
Delay and packet loss measurements are typically performed over a longer period of time in order to allow for the time varying nature of network performance in packet-switched networks.
Add comment
P85
However, the principle of running measurements of long duration conflicts with the crowdsourced user-initiated measurement concept. End users will not accept an extended waiting time for the presentation of results.
Add comment
P86
Conversely, short duration tests can only provide an indication of whether the measurement was done during stable network conditions; whereas long measurement intervals are preferred for the meaningful measurement of the stability of the IAS.
Add comment
P87
While long duration tests are preferred for delay and packet loss measurements, this likely introduces the need for a measurement client running permanently in the background.
Did you know you can vote on comments? You can also reply directly to people's comments.