perfSONAR User Q&A and Other Discussion

[Alan Whinery <>]

On 4/8/2014 3:43 AM, Andrew Lake wrote:
> Hi,
> 
> A negative latency value means NTP is not synchronized. It does NOT
> mean latency is under a millisecond by itself. In real terms, it
> means a packet arrived to the destination before it was sent, which
> isn't possible unless the clocks are wrong. I would recommend
> correcting the clocks if at all possible rather than changing the
> thresholds.

A negative latency value does not mean that NTP is not synchronized.

When the actual path latency is similar in magnitude to typical NTP
drift, synchronized clocks can show negative latency even though NTP is
properly synchronized. Between two hosts on a LAN, where latency is less
than a millisecond, and NTP is synchronized with only network peers (no
hardware clocks), negative latency results in owamp tests will probably
be unavoidable, because clock drift will be on the order of at least a
millisecond. It may be that temperature control and mature filters can
do better, but the way to ensure the necessary clock accuracy for
one-way latency on ~ 1 ms paths is to add hardware clocks to each node.

it looks like Jianan already understood this, but you may have missed
his meaning of "on a LAN".

> 
> That being said, the thresholds in maddash are completely
> configurable. Under the checks section you can adjust the "command"
> option with the thresholds you want. Changing the -c and -w options
> to something containg the negative values will do what you want. For
> example -c "-15:15" -w "-10:10" will make it red if it falls outside
> the range -15 ms to 15 ms, yellow if its outside the range -10 ms to
> 10 ms and green if it is in between -10ms and 10 ms. The notations is
> standard nagios notation so you can do other fancy things with the
> ranges if you are familiar with that but the above should do the
> trick.
> 
> Hope that helps, Andy
> 
> 
> On Apr 7, 2014, at 1:05 PM, Jianan Wang 
> <>
>  wrote:
> 
>> Hi Everyone,
>> 
>> I have a question about having Maddash display alarming red colour
>> while the latency value is minus. This maddash is operated on a
>> LAN, so there could be some minus latency issue from NTP
>> synchronization precision. Apparently the minus latency means the
>> latency is under milisecond, which should indicate the network is
>> pretty good, so I just wonder if I could tweak the Maddash to have
>> latency values lower than zero display green instead of red.
>> 
>> 
>> Thanks.
>> 
>> -- Jianan Wang Department of Electrical and Computer Engineering,
>> Duke University Department of Optical Engineering, ZJU  Information
>> Engineering Qizhen Leadership School, ZJU Tel:(1) 919-937-4582 
>> Email: 
>> 
>> 
>