[ntp:questions] Is a doubling of I/O bytes expected for NTP over Wi-Fi?

Rob nomail at example.com
Tue Nov 26 12:17:47 UTC 2013

David Taylor <david-taylor at blueyonder.co.uk.invalid> wrote:
> Brief background.  I have five Raspberry Pi NTP servers, all with a 
> nominally identical configuration, all with GPS/PPS receivers.  Only one 
> of these is doing any significant work other than running NTP.  None of 
> the devices are acting as servers for other clients.  Two are LAN 
> attached, and three are connected over Wi-Fi.  They all have three LAN 
> servers polled at 32s intervals, and four pool servers (WAN) polled at 
> 1024s intervals.  The I/O shows that input considerably exceeds output.
> Looking at the I/O bytes reported by MRTG for the LAN or Wi-Fi over USB 
> adapters, I see:
> raspi-1  LAN  in: 207 B/s, out: 35 B/s
> raspi-2  LAN  in: 202 B/s, out: 36 B/s
> raspi-4: Wi-Fi  in: 406 B/s, out 40 B/s
> raspi-5: Wi-Fi  in: 426 B/s, out: 53 B/s
> It puzzles me that the bytes sent to the RPi when Wi-Fi is attached are 
> about double the bytes sent over the LAN.  I would expect most of that 
> traffic to be the RPi interrogating the three local stratum-1 NTP servers.
> I don't see why there should be double the bandwidth just because Wi-Fi 
> is used.  Is this expected?  Could it be that the USB Wi-Fi adapter is 
> reporting incorrectly?
> I ask here, as most of the traffic will be NTP packets (or it /should/ be!).

On the receive side there is also broadcast traffic.  This includes
both the usual and subnet-broadcast traffic and the ARP
requests, DHCP requests, etc seen on the network.

On a larger network this can be considerably higher than what you see.

More information about the questions mailing list