perfSONAR User Q&A and Other Discussion

[Terrence Martin <>]

That sounds reasonable. I did manage to find the sonar buoy and get it to 
collect some data. We defintely need to work out the actual perf sonar 
installs though and appropriate testing points. For example in the pingER 
config the default end point is my web server which is not even on my Cluster 
network. Perhaps that information was farmed from the OIM or other source. A 
CE or SE would have probably been a better choice, but if BDII or OIM is 
being used as the source we probably want to publish the correct end point in 
those systems. 

Terrence


On Monday 13 July 2009 12:42:22 Joe Metzger wrote:
> Terrence,
> While the perfSONAR tools & protocols are still under development, they
> can provide significant value to you now.  We have been working  
> extensively
> with the US Atlas community, which has installed perfSONAR tools at  
> almost
> every US Atlas Tier 2. The Atlas administrators have been successfully  
> using
> it to find and resolve network bottlenecks, or to rule out the network  
> as
> the source of some file transfer problems.
> 
> A similar deployment effort is also underway at all the LHC Tier 1's  
> around the world.
> 
> You will want to setup your perfSONAR system to regularly measure the
> network paths between your system and the other CMS Tier 1's and Tier  
> 2's
> that you collaborate with, or depend on for your mission.
> 
> I think it makes sense to try and get together a CMS group to
> recommend basic configuration parameters & test frequencies, and to
> provide a first level of support to the smaller CMS groups in the
> US.  I will talk to Phil Demar & Maxim Grigoriev @ FERMI and see if
> we can get something organized.
> 
> Does this sound like a reasonable approach to you?
> 
> --Joe
> 
> 
> 
> 
> On Jul 13, 2009, at 2:11 PM, Terrence Martin wrote:
> 
> > So as requested in the LHC/USCMS I have perfsonar installed on our  
> > cluster in a spare node. It is still not clear to me, what is the  
> > value to me, as an administrator of a CMS T2 site of having this  
> > service? I have looked through the web interfaces, and aside from  
> > being fairly cryptic and offering several uninteresting tests  
> > between my browser and the node I do not see much else.
> >
> > Since I do not really care about my browser on my desktop on the  
> > 100Mbps building network and really care about my cluster and its  
> > two 10Gbps uplinks (one to ESNet) how do I use these system to  
> > improve the reliability and performance of my cluster? For example  
> > does this node upload data on a constant basis to a central site?  
> > Does it report information like CMS SAM? Does it run tests  
> > periodically between it and other nodes to develop a map of the  
> > current weather of the network? The Pinger service seems to be  
> > capable of collecting some sort of network data, and I saw some  
> > graphs when searching online, but aside from seeing UCSD-T2 in the  
> > configuration it is not clear to me how to actually get anything out.
> >
> > For example, what URL(s) do I enter in the PingER GUI?
> >
> > Terrence
> >
> >
> >
> >
> >
> >
> 
> Joe Metzger
> 
> 
> 
> 
> 
>