[time-nuts] Simple NTP server based on a Raspberry Pi

Peter Bell bell.peter at gmail.com
Mon Oct 29 23:48:48 EDT 2012


When I was playing around with the Raspberry PI, one of the things I
noticed that was the networking seemed to behave rather strangely at
times - not bad enough that you could really say it was broken, but
enough to reduce your level of confidence.  Since I have previously
used that SMSC USB<->Ethernet bridge chip with no problems, I suspect
it's associated with the USB drivers on the PI.

Regards,

Pete


On Tue, Oct 30, 2012 at 7:34 AM, Shaun Kelly <shaun at impsec.net> wrote:
> On 10/29/12 19:17 , Sarah White wrote:
>>
>> On 10/29/2012 7:53 PM, Chris Albertson wrote:
>>>
>>> I'm curious what load average numbers do you get if you type "uptime"
>>> after running NTP for some hours.
>
> [snips]
>
>>> Basically, I'm thinking this method (raspi + refclock) is a good way to
>>> feed a handfull of public stratum 2 servers which by definition don't
>>> have their own local refclock.
>>>
>>> If you made your raspi stratum 1 server publicly available via
>>> pool.ntp.org it would almost certainly be overloaded.
>
>
> I have a Raspberry Pi in the pool now (ntp.impsec.net), though I've set
> bandwidth low (512k) so I am not seeing a great deal of load:
>
> root at ntp2:/home/pi# uptime
>  19:19:57 up 9 days, 21:54,  1 user,  load average: 0.03, 0.08, 0.15
>
>
> top - 19:20:10 up 9 days, 21:54,  1 user,  load average: 0.03, 0.08, 0.15
> Tasks:  55 total,   1 running,  54 sleeping,   0 stopped,   0 zombie
> %Cpu(s):  0.3 us,  0.2 sy,  0.0 ni, 93.9 id,  5.4 wa,  0.0 hi, 0.2 si,  0.0
> st
> KiB Mem:    220652 total,   205332 used,    15320 free,    35256 buffers
> KiB Swap:   102396 total,      880 used,   101516 free,   145376 cached
>
> Watching with tcpdump I am seeing only a few packets per second average,
> withg some periods of burstiness where I see maybe 30 packets per second
> where it will maybe get up to 10-15% busy.
>
> [more snips]
>
>>>
>>> Let's assume a raspi can handle 200 packets per second without loosing
>>> performance too badly (limited by cpu load, etc.)
>
> I'd think that's probably close based on what I'm seeing, though again I'm
> just eyeballing tcpdump.  Mind you, I'm not using the supported turbo modes
> so there might be some wiggle room with mild overclocking.
>
>
> -Shaun
> (sig pending)
>
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.




More information about the time-nuts mailing list