Skip to Content.
Sympa Menu

perfsonar-user - Re: [perfsonar-user] OWAMP Powstream Results

Subject: perfSONAR User Q&A and Other Discussion

List archive

Re: [perfsonar-user] OWAMP Powstream Results


Chronological Thread 
  • From: Andy Lake <>
  • To: Murilo Vetter <>
  • Cc:
  • Subject: Re: [perfsonar-user] OWAMP Powstream Results
  • Date: Fri, 15 Mar 2013 11:00:51 -0400
  • Authentication-results: sfpop-ironport07.merit.edu; dkim=pass (signature verified)

Hi Murilo,

The pattern you are seeing where a large percentage fall very close to one value and there are a handful of outliers is pretty common with owamp data. Generally we attribute the outliers to general background host activity. In the example it looks like greater than 99% of the data is one value. Data like this is why a lot of jitter measurements use the 95th percentile or similar to measure jitter (using the 99th percentile with your example, the jitter would still be 0). I've even heard of proprietary tools that strip out values that fall out of a certain percentile range automatically from reports.  Actually we used to display max values by default on the OWAMP graphs on the toolkit, but switched it to a checkbox option because they were often misleading and not indicative of a real problem. I would say your results look consistent with what most "stable" owamp data looks like and what you are seeing are artifacts of taking a relatively large sample of measurements constantly over time. 

Thanks,
Andy


On Mar 15, 2013, at 10:10 AM, Murilo Vetter <> wrote:

Hello,
 
Here in Brazil, we are concerned about the owamp powstream results. Maybe it was discussed in older threads, but it is still unclear to me.
We are running owamp postream between our measurement points, in a 10 packets/sec rate and summatization in 5 min.
The results that it shows in the measurements are having a strange behavior. A little packets are far away from most of them.
We are suspecting that this can be related to the host usage or something like that, but we are not sure to affirm this.
Could it be some I/O problem? Does anyone already notice something like this?
 
A sample of the measurements is below.
 
<nmwg:message id="message1" type="MeasurementArchiveStoreRequest" xmlns:nmwg="http://ggf.org/ns/nmwg/base/2.0/" xml
ns:nmwgr="http://ggf.org/ns/nmwg/result/2.0/" xmlns:nmtm="http://ggf.org/ns/nmwg/time/2.0/" xmlns:powstream="http:/
        <nmwg:metadata id="meta1">
                <powstream:subject id="subject">
                        <nmtl4:endPointPair>
                                <nmtl4:endPoint role="src">
                                        <nmtl4:address value="mp2.pop-sc.rnp.br" type="hostname"/>
                                </nmtl4:endPoint>
                                <nmtl4:endPoint role="dst">
                                        <nmtl4:address value="mp2.pop-ms.rnp.br" type="hostname"/>
                                </nmtl4:endPoint>
                        </nmtl4:endPointPair>
                </powstream:subject>
                <powstream:parameters id="parameters">
                        <nmwg:parameter name="count">3000</nmwg:parameter>
                </powstream:parameters>
        </nmwg:metadata>
        <nmwg:data id="data1" metadataIdRef="meta1">
                <powstream:datum sent="3000" sync="1" maxError="2.01166e-06" duplicates="0" loss="0" minDelay="0.01
92933" maxDelay="0.117795" timeType="unix" timeValue="1363355702992">
                        <nmtm:time type="unix">
                                <nmtm:start type="unix" value="15343100614190632261"/>
                                <nmtm:end type="unix" value="15343101914118729689"/>
                        </nmtm:time>
                        <powstream:value_buckets>
                                <powstream:value_bucket count="1" value="525312"/>
                                <powstream:value_bucket count="1" value="265216"/>
                                <powstream:value_bucket count="1" value="294912"/>
                                <powstream:value_bucket count="1" value="300032"/>
                                <powstream:value_bucket count="1" value="303104"/>
                                <powstream:value_bucket count="1" value="404480"/>
                                <powstream:value_bucket count="1" value="413696"/>
                                <powstream:value_bucket count="1" value="1206272"/>
                                <powstream:value_bucket count="1" value="433152"/>
                                <powstream:value_bucket count="5" value="197632"/>
                                <powstream:value_bucket count="2979" value="198656"/>
                                <powstream:value_bucket count="5" value="199680"/>
                                <powstream:value_bucket count="1" value="200704"/>
                                <powstream:value_bucket count="1" value="207872"/>
                        </powstream:value_buckets>
                        <powstream:TTL_buckets>
                                <powstream:TTL_bucket count="184" ttl="249"/>
                        </powstream:TTL_buckets>
                        <powstream:reordering_buckets>
                                <powstream:reordering_bucket order="1" count="0"/>
                        </powstream:reordering_buckets>
                </powstream:datum>
        </nmwg:data>
</nmwg:message>
 
Thanks in advance,
-- 
<image001.jpg>

Murilo Vetter
Analista REMEP-FLN/PoP-SC 
PoP-SC - Ponto de Presença da RNP em Santa Catarina 
REMEP-FLN - Rede Metropolitana de Educação e Pesquisa da Região de Florianópolis 
RNP - Rede Nacional de Ensino e Pesquisa 
UFSC - Universidade Federal de Santa Catarina 
SETIC - Superintendência de Governança Eletrônica e Tecnologia da Informação e Comunicação 
Fundações de Apoio FEESC e FEPESE 
--/-- 
www.pop-sc.rnp.br 
remep.pop-sc.rnp.br 
CHAMADOS:  .br 
+55(48)3721-6335 
INOC: 11242*100




Archive powered by MHonArc 2.6.16.

Top of Page