Skip to Content.
Sympa Menu

perfsonar-user - Re: [perfsonar-user] perfSONAR first time use

Subject: perfSONAR User Q&A and Other Discussion

List archive

Re: [perfsonar-user] perfSONAR first time use


Chronological Thread 
  • From: Eli Dart <>
  • To:
  • Cc:
  • Subject: Re: [perfsonar-user] perfSONAR first time use
  • Date: Fri, 20 Jul 2018 10:08:54 -0700
  • Ironport-phdr: 9a23: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

Hi Robert,

Where to put perfSONAR hosts depends a lot on the architecture of your network, where your data flows are, and what kind of testing you'd like to do.

There are some guidelines for the placement of perfSONAR hosts here:

A bunch of the following is covered in that web page, but here are a couple of quick thoughts:
* short-distance throughput tests don't tend to be valuable, and can cause problems. For example, if you set up throughput tests between perfSONAR hosts on either side of your perimeter firewall, the most likely outcome is that you'll clog up the firewall without gaining any insight into the behavior of long-distance data transfers.

* On the other hand, testing from your site to a distant perfSONAR host using both a local perfSONAR host inside your firewall and a local perfSONAR host outside your firewall can help characterize the performance impact of your firewall for long-distance transfers.

* Packet loss monitoring (OWAMP testing) is quite useful both inside a LAN and between a site and the outside world. This doesn't cause congestion problems like short distance TCP tests can cause, and can provide valuable insights into the loss behavior of different parts of your network.

* If you run throughput (pScheduler/iperf) and latency/loss (OWAMP) tests on the same test host and on the same host interface, the throughput tests can cause variation or loss on the host interface itself. This can lead to confusion about the loss data shown in the OWAMP tests, and reduce the utility of OWAMP as an indicator for problems in your network.

Hope this helps,

Eli




On Thu, Jul 19, 2018 at 7:35 AM Roberto Carna <> wrote:
Dear people, I'm new at perfSONAR and I want to ask you a question:

In our company we have two interconnected sites by a L2L link, both
sites are connected to Internet through their own links, and from each
site we have LAN and DMZ's networks.

Because I'm new at this network measurement solution, I need your
opinion about the minimal points from my network scenario where I
should install perfSONAR in order to start analyzing network traffic
and knowing the features we can use.

After the first period of test and understanding, I suppose I will be
in conditions to use perfSONAR in a deep way, deploying it in more
places.

Thanking in advance.

Robert


--

Eli Dart, Network Engineer                          NOC: (510) 486-7600
ESnet Science Engagement Group                           (800) 333-7638
Lawrence Berkeley National Laboratory 



Archive powered by MHonArc 2.6.19.

Top of Page