Interested in learning how you can save power in computer systems, especially with latency critical applications? Learn more about how profile-guided frequency scaling can help solve this problem, with research supervised by Assistant Professor Vinicius Petrucci at the University of Pittsburgh and presented last month at IEEE/ACM CCGrid!

Join us for an introduction to CHI@Edge, Chameleon's foray into supporting experiments with edge devices, what features are implemented, and which are on the way!

Dates, details, and registration at https://chameleoncloud.org/learn/webinars

This month, we announce the early user period for our "edgy" users out there, remind you to please update your list of publications using Chameleon(!) and also let you know of a few important and cool things going on. Stay tuned next month for some announcements about new hardware installations, if all goes well!

As part of Phase 3, we’re looking to improve the Chameleon experience and make the system generally better over the next 4 years. This is your chance to guide hardware purchases, influence Chameleon capabilities and change priorities in our roadmap. 

To help us in this effort, we would like to invite you to take this 10-15 minute survey, and share your experiences, desires, pain points and anything else to help us build a better Chameleon for you! The survey will be open till the end of June. 

Thank you in advance for your participation!

The survey is available here: 2021 Chameleon User Survey

Image from San Diego Law Library

Organizing a summer school, bootcamp or workshop this summer? Having trouble finding a consistent and predictable environment? Learn all about how to use the Chameleon JupyterHub artifact to configure resources (including GPUs!) and create a shared Jupyter notebook computing environment that any event attendee can use. You can manage users, deploy Chameleon resources, and use the Help Desk throughout your event!

This work is part of George Mason University PhD student Zheng Chai and Prof. Yue Cheng’s research on solving federated learning (FL) bottlenecks for edge devices. Learn more about the authors, their research, and their novel FL training system, FedAT which already has impressive results, improving prediction performance by up to 21.09% and reducing communication cost by up to 8.5 times compared to state-of-the-art FL systems.

The summer is almost in full swing, and we're frantically getting stuff done so we can kick back and just rest on our laurels from May until August… just kidding. Expect lots more fun stuff to be announced this summer! For now, enjoy your latest changelog.

In this issue, learn more about running a workshop on Chameleon, blocking out portions of time on your CHI-in-a-Box associate site to "set it aside" for other uses, try out a new Jupyter tutorial around stitching, and apply to take ownership of some of our legacy hardware! We're also very excited to get some comment on IoT hardware investment. More inside!

Image credit: Argonne National Laboratory

Announcing the Chameleon Legacy Hardware Program! We're offering our retired hardware to members of the community interested in joining the Chameleon family as operators of a Chameleon Associate Site. Learn all about the Chameleon Legacy Hardware Program and apply for servers by submitting a short application. Applications are encouraged by 05/16.

This blog feature explores 4th year University of Michigan PhD student Peifeng Yu’s research on hyperparameter tuning, presented earlier this month at MLSys21. Learn more about Yu, the hyperparameter tuning engine, and how it can improve your deep learning model training process.

Do you have a research cluster with an inflexible interface? Does it not provide the level of access (root) you need to run repeatable experiments? Would you like a well-defined and easy way to contribute under-used resources to the community? Learn what it takes to set up a Chameleon associate site which will give you access to industry-standard APIs, best practices and automation to run a research testbed, and support from the Chameleon team for you and your users.