Great news in Chameleon-land!
The summer is heating up--at least in Chicago--and here we are working on making the system better so you don’t have to! (We suggest using any newfound time floating in a pool in a large duck-resembling inner tube whilst sipping a cool drink.)
CHI-in-a-Box beta release out now! Last December we announced the second alpha release of CHI-in-a-Box. Since then, we have dramatically improved how Chameleon works under the hood, first by moving to a container-based infrastructure, then by upgrading the entire system to the latest OpenStack releases. In the process we looked at CHI-in-a-Box and decided to change things up: instead of releasing and maintaining a separate packaged version of Chameleon independently, we decided to open-source our own deployment. This expands the scope of CHI-in-a-Box to include additional monitoring, maintenance tools, operational runbooks, and more. Our lessons learned can now be properly passed on to you! We hope that this improves the long-term health of CHI-in-a-Box as a program and makes it easier for you to consider Chameleon at your own institution. Have a look at the official GitHub repository, and contact us with any questions.
New MPICH appliance available for all experimenters. Good news for parallel computing researchers! The MPICH development team at Argonne National Laboratory contributed a new appliance with the latest version of MPICH (3.3.1) installed. MPICH is a high-performance and widely portable implementation of the Message Passing Interface (MPI) standard. To learn more about MPICH, please visit the official website. This appliance is built on top of the latest version of CC-CentOS7, so it has all Chameleon utilities installed, such as cc-snapshot, cc-cloudfuse, and the metrics collection agent.
Ubuntu18.04? Yes. CUDA10? Also yes. Hot on the heels of our announcement last month of a new Ubuntu18.04 appliance, this month we released a new CUDA 10.1 appliance for Ubuntu18.04. This appliance can be launched on the GPU-enabled hardware at CHI@TACC.
Support for updating complex reservations. When we expanded the Chameleon reservation system earlier this year to include Floating IPs and VLANs, we made it possible for you to reserve all the resources required for an experiment in one lease (and we updated our GUI to prove it!) One thing was still ahead of us though: allowing updates to complex reservations, such as changing the amount of Floating IPs needed, or requesting a different VLAN, or adjusting the amount of nodes in one of these complex leases. That limitation is now gone, enabling you to have more flexibility when making your leases. We encourage you to try out the new reservation types, in particular Floating IPs, which are relevant to virtually all types of experiments. Check out the documentation to learn more.
Faster response to allocation modifications. In the past, there would be some delays between adding a user to a project and them being able to actually access Chameleon, because the change took a bit to roll out to all of our systems. We took a look at this long-standing issue this month and have made some changes to speed this up significantly, which should reduce problems you (or, perhaps students in a class you’re teaching) might have with not being able to log in after project members are updated. If you find you ever do have questions about your allocation, first check our FAQ to see if it answers your question, but always feel free to send us a ticket on our Help Desk - that’s what it’s there for.
Finally, we will be having another Introduction to Chameleon webinar this month on August 27th. This is a great opportunity for colleagues or students perhaps not familiar yet with Chameleon to learn the basics, but might also be a good way to learn some easy-to-use features that you aren’t already aware of. Sign up here!