Great news in Chameleon-land!

 

We have a few goodies below to start off the New Year on the right foot!

 

You may have noticed that our attention has been slowly shifting to upgrading and improving the system rather than extending it: we are not announcing new features as aggressively as at the beginning of phase 2 – but then we noticed that users often ask for features that have been in the system for a while or are not aware of tools like orchestration that can make their experimentation easier and more repeatable. Therefore, for the remainder of phase 3 we will shift focus onto “digesting” the capabilities we have, i.e., making them better, more reliable, and easier to use – but also better known via webinars and other outreach activities. This being the case, we will suspend the changelog posts for now and instead announce new features and upgrades as well as training opportunities via the users@chameleoncloud.org mailing list as they come up. Enjoy the new things that we did in January!

 

Simpler SDN experiments at UC. Before, if you wanted to utilize SDNs in your experiment, you were limited to using the Skylake nodes, as they were the only nodes logically connected to a switch capable of creating isolated OpenFlow-ready networks (our Corsa DP2200/DP2400s). This limited the scale of your networking experiments. Now, that limitation is lifted: you can now additionally include Haswell nodes at CHI@UC participate in SDN experiments! We have updated the documentation to include the OpenFlow port mappings for each Haswell node. If you haven't already played with SDN and the BYOC feature, perhaps now is the time? Check out this example tutorial (it will launch in the Chameleon Jupyter environment!)

 

New Jupyter tutorial notebook. Speaking of Jupyter tutorials, we have another one, which devoted blog followers may have already taken notice of. This tutorial explores how you can orchestrate a distributed system and configure each node's DNS resolution (using /etc/hosts) such that you don't need to hard-code IP addresses in your experimental configuration: an important property for reproducibility (and for maintaining your own sanity). It also covers reserving multiple nodes, IP addresses, and creating isolated networks, all within Jupyter.

 

KVM-2015 officially decommissioned. As announced, we have disabled the old KVM cloud in lieu of its more modern replacement. Please send us a message via a support ticket if you find you need assistance with setting up your experiments on the new KVM cluster.

 

System upgrades in February. We'll be performing some regular updates to the testbed in February, notably upgrading to the latest OpenStack installation (Train), which includes a number of usability and performance improvements, as well as bug fixes.

 

Upcoming webinars

• Feb. 11th @ 2-3pm CT | Orchestration: Ansible
• Mar. 10th @ 2-3pm CT | Orchestration: Jupyter Notebooks
• Apr. 7th @ 2-3pm CT | Intro to Chameleon
• May 12th @ 2-3pm CT | Orchestration: Heat Templates

 

Conference submission deadlines

• TaPP & IPAW (Mar. 1st)
• ECCV (Mar. 1st)
• VLDB (Mar. 1st)
• IEEE CDC (Mar. 3rd)
• ACM SIGIR (Mar. 5th)
• ACM ESEC/FSE (Mar. 5th)