eduPERT Reports

27.02.2014: APM Discussion

EduPERT February call
Roundtable about APM and the possible role of eduPERT.
Brian Nisbet gave a short introduction about the topic and the slides can be found
in the section Documents.
Here is the links to some tools that were posted during the call

Application Performance Management/Monitoring - Brian Nisbet HEAnet

"Increasingly NREN clients are outsourcing their applications and
learning environments to 3rd parties outside the GEANT network. While
simple measurements to these locations are relatively easy, ping does
not measure how quickly a user can login or how the application
performs. HEAnet are currently investigating ways of providing
Application Performance Management (APM) services to our clients. In
addition, we are discussing the role of the NREN (and groups like
EduPERT and the GEANT MDSD) in regards to the performance monitoring &
management of these 3rd party applications."

27.01.2014: Science DMZ

EduPERT January call: Happy new year!
Following the abstract of the very interesting presentation we had and
the slides can be found in the section Documents

Science DMZ - Jason Zurawski ESnet

"Big Data", a recent area of emphasis meant to address the growing
research capability of scientific disciplines, has caught many off
guard. The exponential growth patterns, spanning areas from Physics to
Medical research, stand to overwhelm the capabilities of many legacy
infrastructures in the coming years. Network providers have a unique
challenge in this space: how can the networks of today scale to meet the
coming data demands, while remaining both safe and reliable? The Science
DMZ network design paradigm combines 3 core concepts to deliver on the
requirements of scientific network use:

- A stable and robust architecture to support data intensive applications
- Capable couplings of hardware and software to migrate and process data
- A robust framework of verification and testing tools to ensure long
term health and reliability

This talk will outline the challenges in supporting data intensive
science as identified by ESnet, and will highlight design and
implementation best practices that are currently used around the world
to best support the needs of operators and end users.

25.11.2013: EC Internet Monitoring Study

EduPERT November call
Timur Friedman, professor at Sorbonne University gave an overview of the internet
monitoring study organized by the EC.
Timur is one of the responsible people for the study and he kindly
accepted to participate and present at our call. You can read more about the study on
the European Commission link

28.10.2013: MonIPÊ Project

EduPERT October call
Alex Moura from RNP presented the MonIPÊ Project.
This Brazilian project is about building a measurement service using small and low cost
hardware like Rasperry Pi and Cuboxes.
PerfSONAR will be installed on the boxes and shipped to the end-users (Campuses and
Extremely interesting project, you can read more about it and the measurement tests that
will be done on the slides that are published in the section Documents on this portal

24.09.2013: Tracebox and EC workshop

Here is the recap of the eduPERT September call:

- Presentation from Gregory Detal:
Gregory Detal is a researcher at Université catholique de Louvain in
Belgium. He works together with Oliver Bonaventure on MPTCP, but during
this call he presented his work on a new tool able to discover the
presence of middleboxes on the path and their interference. The name of
the presentation was "Detecting middlebox interference with tracebox"
and here are the slides:


Extremely interesting work that can be very useful when troubleshooting
performance issues. Still many thanks to Gregory.

- The second part of the call was a round table to answer some of the
key questions for the EC workshop that will take place on the 3rd and
4th of October in Brussels.
Main points coming from the discussion:
1. Effort of GEANT towards more data sharing (making publicly accessible
the perfSONAR archive is one possibility)
2. A tool that visualize the path together with the results of the
measurement is missing (a sort of integration of traceroute in perfSONAR
is already under discussion). Even better would be a tool able to show
the intermediate performance for each hop on the path and at the end
host as well.
3. The tool that Gregory presented coupled with other measurement tools
would help noc and pert engineers in having an easier life.

21.08.2013: MPTCP and traceroute

During this call we had a presentation and a short demo.

The presentation was given by Benno Overeinder from NLNlabs on MPTCP.
He gave an overview of the protocol explaining how it works and which
are the major problems when the protocol is used in real environments.
The slides can be found on the portal:

There is also a new page on our KBpert about MPTCP

During the second part a demo was given by Robert Stoy on my traceroute.
The aim of the demo was to demonstrate that sometimes mtr can show loss
on the network where there is actually none due to rate limiters,
timeouts and cpu cycles sparing.

19.06.2013: virtual F2F meeting
This call was about the TNC F2F meeting, its outcome and what we would like to do in the next future.
The discussion was very alive and in the end we all agreed on how to proceed.
Here is a brief report of the call.
The main two points that were raised during the TNC F2F meeting were:
Outreach and transatlantic troubleshooting.
A first discussion about them took already place during the F2F in Maastricht but things were better defined during the call.
The first decision we all agreed on was that we will focus more in doing outreach especially towards end user communities.
This is something we have already started. In September eduPERT will hold a training at the EGI technical forum.
Next possible communities are Eudat and PRACE.
Second decision: we will not consider to do transatlantic troubleshooting at the moment because of
lack of resources and because troubleshooting is in general a task of the local PERT team involved in the performance issue.

23.05.2013: 2nd Virtual Workshop
In this workshop Robert Stoy showed us how to install and use bwctl with iperf.
The config files and a  text file containing all the commands he used during the
workshop can be found in the documents.

15.4.2013: April eduPERT phone call
The main topic of the call was Happy eyeballs. The reason behind this
choice was the idea of writing a new page in our KB as outcome of the call.
So you will read the report of this part of the call in the page that
will be written :)
In the meanwhile you can check this Keynote speech [1] that Geoff Huston
gave last year at NORDUnet 2012.
The presentation is very well done and it is not just about happy
eyeballs but more around the entire concept of dual stack.

The rest of the call was about eduPERT and SA4 in GN3+.
Last week the kick-off of the activity took place. These are the tasks
in SA4 (Network Support Services):
Task 0 - Activity Leadership: More related to the management of the
Task 1 - Multi-Domain Monitoring (MDM): Basically PerfSONAR.
Task 2 - Multi-Domain Security Services (MDSS): Design, implement and
support multi-domain security services
Task 3 - End-to-End Management – Network Management Architecture
Approach: The main objective of this Task is to enhance the design and
use of existing monitoring tools and systems by aligning them with
business processes
Task 4 - eduPERT: Us.
Task 5 - GEMBus Pilot: The pilot will last 10 months at the end of which
GEMBus will become a service or will be cut.

During the kick-off every task leader gave a presentation about the
current status of his task and the plan for the next year(s)
For eduPERT current Status there are the points that I presented:
- A small but steady core group
- A bigger community that also includes end-users
- Monthly phone calls
- F2F meetings
- eduPERT Portal
- Performance U!
- Winter School in March: A pilot that was a success

And for the next future:
- Strengthen all the points of the previous slides
- Invest more on Performance U! and improve the overall structure
- Move towards communities of end-users more than try to reach the
single one.
- Invest more internationally (already on going)
- New pilot for the summer school: EGI

General announcement: eduPERT F2F meeting at TNC2013 [2]


27.2.2013: February eduPERT phone call

It was fun to discover the different uses and options that nuttcp provides.
But before I start with the report I have an announcement.
Considering the success of this call, we will have more sessions like this in the future. Bartek had this nice idea and he named it "virtual workshop".
The virtual workshop will take place once every 2/3 months instead of the normal call. We will use tools or software and try to understand it together. Bartek will be in charge of it.
Any topic coming from the community is more than welcome, you only have to send him an email with details.

Now back to the today's call!

Here are the different options we tried using a server here at SWITCH and a server in Poznan.

- traceroute
nuttcp -xt host@poznan

-simple initial test
nuttcp -4 -i 1 host@poznan

-reverse traffic (from host@poznan to host@switch)
nuttcp -4 -r -i1 -T20 host@poznan

- "Setting the "-s" option causes nuttcp to write its data to standard output rather than using prefabricated memory buffers. This option is disabled for security reasons on the server. It works only with TCP and with one single shot server".
This time we used my mac to send the traffic.
On the main host
nuttcp -1 -s --nofork
On my mac
nuttcp -i1 host@switch

what we saw was actually a stream of character on standard output of the host here at switch.

-"-N specifies the number of parallel TCP (or UDP) data streams to be used for the data transfer, with the default being a single data stream. The maximum number of parallel data streams that can be used is 128"
nuttcp -4 -N2 -i1 host@poznan

- -R option:"The transmitter rate limit throttles the speed at which the transmitter sends data to the network, and by default is in Kbps, although the 'm' or 'g' suffix may be used to specify Mbps or Gbps." This option may be used with either TCP or UDP data streams
nuttcp -u -i 1 -R10m Host@poznan

-"instantaneous rate limit with optional packet burst"
nuttcp -u -Ri300m/20 -i 1 host@poznan

Very interesting and useful feature. It gives the possibility of shaping the traffic.

Third party mode ran from my mac
nuttcp -r -i1 host@switch host@poznan

it's enough to have nuttcp running on two machine and have a third one to decide how to run and shape the traffic and collect statistic.

25.1.2013: January eduPERT phone call

Here are the minutes!

-TF-NOC meeting: Bartek Gajda, Tim Kleefass and I attended the meeting.
The slides of our presentations can be find here

-TIP 2013
-- Many talks on SDN. The most interesting one in my opinion was given
by Ronald van der Pol (who unfortunately could not attend today's call)
MPTCP in a Multipath WAN Fabric using Openflow

--The new perfSonar-ps Dashboard from Esnet
--GeneTorrent: Software for high performance big data transfers
developed at UCSC and used by the CGHub (Cancer Genomics Hub) user
community. An idea based on the bit torrent concept which enables to
have multiple TCP streams from multiple computers to parallelize genome

-Nuttcp: I came across the tool during the Network Performance Workshop
at TIP. Very similar to iperf but with the very good feature of being
stable when used with UDP traffic. We agreed on postponing this topic to
the next call. The plan is to have more time in order to play and test
the tool together setting server and client in different locations in

- Highlights from the talk about Perfsonar given by Domenico Vicinanza
--10 Gb/s perfSONAR boxes will be installed in the next couple of months
in the GÉANT backbone.
--Pre-installed VMs with everything installed.
It is currently available here:
It will be soon available here:
--First perfSONAR MDM Hands On Training
Berlin, Germany, 19-21 March 2013
Description of the event can be found here:
I and Robert Stoy have already registered to join the training.
For the next future there is the idea of using a usb pen for perfSonar
and also of running some performance tests in order to check the
stability of perfSonar in a virtualized environment and I specifically
asked Domenico of involving eduPERT during these tests.

-General announcement:
--More involvement of eduPERT on an international level with other
communities like South America, New Zealand and APAN besides internet2
with whom we have already a very good collaboration.
--Mail from Chris Welti on the pert-discuss about from the
MLab generated a very interesting discussion. Many of us tried to test
the page and got more or less similar results (download ~35Mb/s, upload
~6Mb/. Isn't that too low?) with the exception of IE10 which would give
back for the download something like 8Mb/s (????). Discussion on what
could be improved (it is not possible to choose for example the location
of server to be used for the test) or how much the interface of the page
or other software running on the laptop might influence the test and so on.
Simon Leinen launched the idea then to try to troubleshoot what is going
on behind the scene and send the results to the pert-discuss. You are
more than welcome to try this as well. Go to the page
run your test maybe with wideshark running on the background and using
different browsers.
--Bartek Gajda referred to a survey that TF-NOC has published where
there are also info about network and performance tools that could be of
interest for our community
--Alan Buxey pointed us to a very interesting article about performance
in wifi networks and what parameters influence the wireless throughput

30.11.2012: November eduPERT phone call

Participants: Bartek Gajda, Tim Kleefass, Simon Leinen, Ronald van der Pol, Alessandra Scicchitano,  Robert Stoy.
Apologized: Alex Gall, Chris Welti.


    The next phone call has been fixed for the 25th of January 2013 at 10am CET. There
    won't be any phonecall in December because of the Christmas vacation.