Using One Way Latency with Measurement Framework

[Prateek Jain]

Hi… I am trying to use one-way-latency(owl) in my experiment setup. I am following the notebook KNIT6_owl_mf from jupyter_examples_1.5.4. when I try to execute the following step

from mflib import owl
slice_name="MySlice"
local_owl_dir=f'/home/fabric/work/owl_output/{slice_name}'
owl_mf = owl.OwlMf(local_owl_dir = local_owl_dir)

I get the following error:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[7], line 1
----> 1 owl_mf = owl.OwlMf(local_owl_dir = local_owl_dir)

AttributeError: module 'mflib.owl' has no attribute 'OwlMf'

I am using the fabrictestbed-mflib==1.0.3 because I am using ELK for monitoring my experiments (refer to this post)

• This topic was modified 7 months, 3 weeks ago by Prateek Jain.
• This topic was modified 7 months, 3 weeks ago by Prateek Jain.

[Mami Hayashida]

Yes, this is caused by a rather major change in the mflib.owl code that I made recently.  Please refer to the new version of OWL examples I am preparing for KNIT7 found here (Notice it is still on a feature branch): https://github.com/fabric-testbed/jupyter-examples/tree/mflib-owl/fabric_examples/public_demos/KNIT7/one_way_latency_measurements.  With the mflib version you have posted, these two notebooks should work. The “prep1” notebook is for creating a slice and installing/setting up all the prerequisites on each remote node; “demo1” uses the same slice and runs OWL in 3 different ways.  Please do let me know if you run into any problems.

We will continue to add minor edits to the OWL-related sections of Measurement Framework, mflib, and jupyter examples throughout next week and will have a demo at KNIT7. At that point, the newest version of demo notebooks will also be available.

[Prateek Jain]

Hey Mami.. I checked the link that you shared but I could not find instructions for using owl for MF slice. The link that you provided prepares a non MF slice for owl. My slice is already MF slice so I am not sure if it works for me.

I have another question.. I am not sure if you can help me with that. So in my experiment I am transferring a file from one node(on site 1) to another node(on site 5).. passing the file through some intermediate nodes on site 2, site3 and site 4. I want to calculate the precise time for my file transfer from site 1 to site 5. I am not sure which FABRIC measurement facility I can use to calculate this. It will be of great help if you could guide me in the right direction on how to achieve this (maybe you can share a link to some content about such functionality or some link to jupyter notebook about such functionality will be highly appreciated).

Thank You

[Mami Hayashida]

Prateek,

What is shown in the owl-demo notebook (not the prep one) will work on any slice, whether it is an MF slice or not. All it is doing, basically, is: 1) pulling fabrictestbed-owl image from Docker Hub to each remote node; 2) start docker containers for senders + receivers using the image; 3) download the collected data from the receiver (destination) node(s).  Because your slice has a measurement node, it is possible that 1 or more of the for node in nodes: may have to be edited to exclude the meas-node. (and that may not even be necessary). Let me know if you run into any issues.

The second question: It just so happens that my colleagues from the University of Kentucky will be giving a tutorial at the KNIT7 on PTP timestamp experiment which involves sending a probe packet from one node that will be forwarded through multiple nodes to a destination node while recording the packet arrival time at all intermediary nodes. Yours sounds quite similar, so if you will be at KNIT7 (and can wait till then), you may want to attend that session and talk to the presenters there. In their case, I believe, a packet being “time-traced” is a ping packet while you are sending a file — but you may very well be able to make minor adjustments to their script.  If you won’t be at KNIT7, those notebooks should become available around that time, if not earlier.

If this does not satisfy your needs, however, do let me know. I will need more information, but I am sure, with what we have already implemented, you can run this experiment without much trouble.

[Prateek Jain]

Hi Mami.. I will try the approach that you suggested..

This is exactly what I want to do… I plan to send a tracer packet just before sending the file and after sending the file… and check my file transfer time.. I will be there at KNIT7 and if I could not figure out how to timestamp and check packet then I will definitely be in contact with the team working on it..

Thanks 🙂

[Mami Hayashida]

Sounds good. And I will be at KNIT7 as well, doing an OWL demo. So I’ll probably see you then.

[Prateek Jain]

Hey..

I am trying to access the link that you have provided for the OWL measurement

https://github.com/fabric-testbed/jupyter-examples/tree/mflib-owl/fabric_examples/public_demos/KNIT7/one_way_latency_measurements

But now the link is not available… it says 404 not found..

[Mami Hayashida]

Sorry, my fault!  I deleted that feature branch after it was merged into ‘main’!   Try this: https://github.com/fabric-testbed/jupyter-examples/tree/main/fabric_examples/public_demos/KNIT7/one_way_latency_measurements.

[Author]

Posts