Great news in Chameleon-land!
November is fast upon us, and that means two things: discount candy corn and Chameleon changelogs.
New KVM site online! In 2015, shortly after the release of the first version of the bare metal testbed, we launched the Chameleon KVM cloud as a companion platform for those who were more interested in a more flexible virtualized environment for experimentation. We’re happy to announce that we’ve given our KVM cloud a major facelift and have upgraded it to the recent OpenStack Rocky release. What this means in practice is better reliability, more capabilities, and better interfaces to the testbed. The new cloud is available today at kvm.tacc.chameleoncloud.org. Both the new and old KVM site will be running simultaneously to allow you to migrate workloads over time. We anticipate an end-of-life for the old KVM cloud at January 1st, 2020, at which point the site will be taken offline. We have already migrated most of your images, key pairs, and other data tied to your project to the new environment (as of October 31), so you shouldn’t have to do any heavy lifting! Please let us know how you like it, and as always, you can checkout the documentation here.
Cloud trace data now available for bare metal sites. We have at various times released anonymized usage data on the Chameleon KVM cloud for use in simulations or analysis. This month, we have updated both our open-source trace generator and the trace format specification to serve for both the virtual machine and bare-metal clouds. We also published a new KVM trace from our KVM site and bare-metal traces from the CHI@UC site and CHI@TACC site using the updated trace format version. For more information about cloud traces, please visit our cloud traces page on Science Clouds and contact us if you would like to either use our data or contribute your own.
Automatically enable jumbo frames for your networking experiments. In order to take full advantage of the 10G NICs in our bare metal nodes, it is generally advised to set a higher MTU on your network devices. However, this requires support across all switches/routers in the network path. We’ve done a thorough run-through of the Chameleon networking infrastructure to ensure this is the case, and in particular paid attention to the 100G link we have between the TACC and UC sites. You can rest assured that your packets are able to flow with maximum throughput. To find out how to easily set up your custom network for jumbo frames (and avoid having to set the MTU settings on your nodes manually!), check out a new section of our documentation about this.
Usability improvements in the reservation system. We continue to make improvements to how the reservation UI works when you are reserving bare metal nodes and/or network resources on the testbed. In particular, this month we made sure all the displayed time zones in the lease calendars were consistent to avoid confusion about when a particular resource is available.
New appliances. Finally, we have two new appliances for your enjoyment this month. The first is a deployment of perfSONAR, a network performance monitoring tool that can be set up to continuously monitor end-to-end link performance over time. We’ve also updated the DevStack appliance for the OpenStack Rocky release, allowing you to easily launch an isolated OpenStack installation of your own. Hope you enjoy!
The following most recent webinars are now available for online viewing:
- Nov. 14th @ 2-3pm CT | Chameleon Advanced Topics: Bring Us Your Troubles and We Will Shoot Them!
Submit your questions: bit.ly/nov14questions
- Dec. 10th @ 2-3pm CT | Make the Best Use of Your Allocations: Orchestrate Your Experiments
- Jan. 14th @ 2-3pm CT | Intro to Chameleon