Skip to Content.
Sympa Menu

perfsonar-user - Re: [perfSONAR-developer] [perfsonar-user] Updated Statement on Researcher Participation

Subject: perfSONAR User Q&A and Other Discussion

List archive

Re: [perfSONAR-developer] [perfsonar-user] Updated Statement on Researcher Participation


Chronological Thread 
  • From: Alan Whinery <>
  • To:
  • Subject: Re: [perfSONAR-developer] [perfsonar-user] Updated Statement on Researcher Participation
  • Date: Wed, 21 May 2014 09:41:03 -1000

The RRD SNMP graph shows 5-minute averages of octets counted at an interface, while the perfSONAR toolkit graph shows the average throughput result of 20-second (probably) tests. So the averages will show lower numbers than the test results.  If you ran a 5 or 10 minute bwctl test, you would see a distinct feature on the SNMP graph, but a 20 second test doesn't influence a 300 second average very much.



On 5/21/2014 9:21 AM, Daniel Doyle wrote:
Devine,

Looking at your attached picture I feel that I might not be getting the full picture here. As you said that these are two graphs representing traffic over the same link I do not see how the SNMP monitored one can show less traffic than the BWCTL throughput test (perfSONAR). If anything I believe only the inverse could be true where the SNMP monitored one will simply show all traffic going across that link whereas the perfSONAR one will only show traffic generated from the test, which is a subset of or at most equal to the total link traffic. Before I could weigh in any more it might be helpful to double check that and/or explain a little more about your setup and what you are trying to accomplish and where these measurements are coming from. Of particular use would be to know the sampling interval on each as if the sampling for the SNMP is relatively slow, say every 5 minutes or so, that could account for the difference. Regardless though I believe that assuming they both utilize the same link the total traffic passed will be indicated on the SNMP counters.

With respect to getting at the raw data in perfSONAR, you could use direct SQL access. Someone else might weigh in here differently but I would probably recommend checking out the SOAP based API for the existing Measurement Archive (MA) as it takes care of some of the more underlying bits in terms of tables and columns and joining it all together for you. I believe a decent starting point might be here: http://psps.perfsonar.net/client-doc.html . It's also worth nothing that there is a new version of the MA that is coming out with a RESTful API that should be easier to use, but that's likely a few months away and may not meet your timing needs.

-Dan

On May 21, 2014, at 7:50 AM, Tshabalala, Devine (Mr) (s210051701) <> wrote:

Thank you Daniel.

Are there any people who are capable and can help out (especially on technical implementation ).
The attached is a screen shot of two different screens monitoring the same link (10 gig link). Screen 1 monitored by SNMP (production traffic) and screen 2 monitored by perfSONAR (throughput being utilized). The two graphs have to be visually compared to determine the total bandwidth being utilized. Hence, l want to integrate these tools to generate a consolidated network performance monitoring report to effectively manage bandwidth utilization. So what can be the challenges that l need to look at when it comes to integrate perfSONAR and other tools? Also l want to know, how the database table formats look like.
Thank you :-)

Kind regards
DevineTshabalala
Masters Student
Department School of ICT : l.T. Communication Networks
(North Campus)
Tel: 078 491 6130

 

A successful man is one who can lay a firm foundation with the bricks others have thrown at him and
Action is the foundational key to all success. 



From: Daniel Doyle <>
Sent: Tuesday, May 20, 2014 10:56 PM
To: Tshabalala, Devine (Mr) (s210051701)
Cc: ; 
Subject: Re: [perfSONAR-developer] [perfsonar-user] Updated Statement on Researcher Participation
 
Hi Devine,

Firstly, thanks for taking an interest in perfSONAR! To answer your question, I don't think anyone is aware of any formal efforts to correlate perfSONAR and SNMP data, though it's certainly possible that there might be informal or side efforts in the community that could be a starting place. I am forwarding this email over to our users list to see if anyone wants to chime in with anything they might be up to that could be a foothold for this.

Certainly feel free to ping either the developer or the user list if you have any other questions.

Thanks,
-Dan

On May 19, 2014, at 8:09 AM, Tshabalala, Devine (Mr) (s210051701) <> wrote:

hi

Hope all is well :-). l am a masters student at Nelson Mandela Metropolitan University in South Africa. l am networks passionate student, wanting to do a research on network monitoring and performance. I just wanted to ask if they has been project that correlate of perfSONAR and SNMP (or similar) data? l have an interest of researching towards that direction? Therefore, l am searching for some hints and guidance on how to tackle such a situation.
You assistance would be greatly appreciated.. 


Kind regards
DevineTshabalala
Masters Student
Department School of ICT : l.T. Communication Networks
(North Campus)
Tel: 078 491 6130

 

A successful man is one who can lay a firm foundation with the bricks others have thrown at him and
Action is the foundational key to all success.

Dan Doyle
GlobalNOC Software Developer
1-812-856-3892



--> <inteer.PNG>

Dan Doyle
GlobalNOC Software Developer
1-812-856-3892







Archive powered by MHonArc 2.6.16.

Top of Page