Chameleon Changelog for May 2019

Great news in Chameleon-land!

It’s summertime, and that means open-flame grills, pools, and hot temperatures: all things that our servers absolutely hate, the poor things. Fortunately they get to stay cool in our data center, and even more fortunately, they even got some nice upgrades.

100G networking at CHI@NU now ready for experimentation. Back in October we announced a new Chameleon Associate Site at Northwestern. The Northwestern site has a few specialized nodes, notably nodes with a Mellanox ConnectX-5 100GE card, allowing for 100G experimentation on a single node. However, up until now, that node wasn’t able to take advantage of these high bandwidths on paths in or out of the testbed. As of this month, these nodes are now connected to 100G-capable uplinks, allowing you to connect the Northwestern site to other Chameleon sites or academic institutions (via ExoGENI stitching). Access to the CHI@NU site is still on an as-needed basis, so if you are interested please open a ticket via the helpdesk to request access and  include “CHI@NU” in the subject of the ticket to help us expedite your request. Once your account on the new site has been activated, you will be able to reach the site at

More disk space on P100 GPU nodes. In response to many user requests, our P100 GPU nodes at TACC have been upgraded from 250GB to 1TB of storage space! We hope that this improved capacity assists you in experimenting with larger datasets and models on these nodes. Side note: did you know Chameleon has over 2PB of internal storage available via the Object Store? It’s a good place to store experimental artifacts you might need closer to your bare metal nodes, like large data sets or trained models. There is a FUSE mount conveniently located on any instance you launch with Chameleon; learn how to use it here.

Provision layer-2 circuits with one command! If you’re already familiar with stitchable VLANs via ExoGENI, you’ve probably been interacting with the GENI tools to provision those links. There can be a lot of steps to this, and it’s been our goal to provide a simpler way to do this. That day has come! There is a binary tool, available in preview in our JupyterHub environment, that allows you to provision circuits with one command using your GENI credentials. You can check it out in our DirectStitch tutorial.

Speaking of tutorials… in March we announced some new tutorials you could take as part of our JupyterHub environment. We’ve added to our set of tutorials substantially in May, in particular covering orchestration, data management (including snapshotting and use of the Object Store), the command-line interface, the Python library interface, as well as advanced networking capabilities like stitching. Check them out when you have a moment! We also freshened up our introduction to Chameleon tutorial, which could be helpful for any students or members of your group who are new to the testbed.

Finally, you should know there will be an Introduction to Chameleon webinar on June 18, 2PM CT. Please pass on this mail to anybody who might be interested in a crash-course in Chameleon! Though you are already a Chameleon user, you might learn some new things yourself. Please register for the webinar here.

Add a comment

No comments