Chameleon, led by Senior Scientist Kate Keahey from Argonne National Laboratory, has been a cornerstone of CS research and education for nearly a decade. The platform has served over 10,000 users, contributing to more than 700 research publications. Chameleon has now secured an additional $12 million in funding from the U.S. National Science Foundation (NSF) to roll out its next four-year phase. With this new funding, Chameleon will continue to innovate and support its growing community, enabling groundbreaking discoveries in CS systems research.
ABOUT CHAMELEON: A PLATFORM FOR INNOVATION
Chameleon is a large-scale, deeply reconfigurable experimental platform built to support CS systems research. The platform enables a wide range of community projects, from developing new operating systems, virtualization methods, and performance variability studies to power management research, software-defined networking, artificial intelligence, and resource management.
To facilitate such diverse experiments, Chameleon provides a bare metal reconfiguration system, granting users full control of the software stack, including root privileges, kernel customization, and console access. While most testbed resources are configured this way, a small portion is configured as a virtualized KVM cloud. This balance allows for both fine-grained resource sharing and the strong isolation properties of bare metal configurations.
Unlike traditional CS experimental systems typically configured by in-house infrastructures, Chameleon leverages OpenStack, a mainstream open-source cloud technology. This adaptation provides practical benefits, including familiar interfaces for users and operators, workforce development potential, contributions from a community with membership over 2,000 strong developers, and the ability to influence infrastructure used by millions of users. Notably, the Chameleon team played a leading part in contributing the OpenStack Blazar component handling advanced reservations.
As an open testbed, Chameleon has supported over 10,000 users working on more than 1,100 projects since its public availability in July 2015. The platform continues to evolve, with hardware and feature additions documented in its news and blog items.
A NEW PHASE
A significant highlight of Phase 4 is the formal integration of the National Center for Atmospheric Research (NCAR) as an official Chameleon site and partner. NCAR joined the Chameleon project as a volunteer site in January 2022 with a donation of ARM Thunder X2 nodes; NCAR’s new status opens up vast opportunities for collaborations in translational computer science. This partnership is expected to drive advancements in edge computing and edge-to-cloud AI, areas where Chameleon has already made substantial investments and gained significant recognition.
“Innovation in computer science takes place at an accelerating rate, so that if we are not efficient at integrating it, increasingly more opportunities are lost”, said Keahey. “Partnership with NCAR allows us to bring together the communities of computer science and atmospheric science researchers to explore a translational approach where problems in atmospheric sciences drive computer science research, and emergent solutions in computer science are directly translated into capabilities for domain science.”
In response to the increasing demand for Artificial Intelligence (AI) and Machine Learning (ML) research capabilities, Chameleon will be enhancing its infrastructure with the acquisition of new GPUs, including H100s arriving at TACC this fall. This upgrade will bolster support for AI/ML research, enabling more efficient and powerful computational experiments.
For instance, researchers working on large-scale neural networks will benefit significantly from the H100 GPUs, which offer improved performance, faster training times, and the ability to handle larger datasets. This enhancement will facilitate advancements in various AI applications, such as natural language processing, computer vision, and autonomous systems, by providing the necessary computational power to develop and test sophisticated models.
Chameleon’s fourth phase will also see continued efforts in edge computing and the Internet of Things (IoT), building on the successful initiatives from previous phases. The platform will further develop CHI@Edge, which extends support to Single Board Computers (SBCs) deployed outside traditional datacenters. This initiative supports a wide array of experiments, from network fingerprinting to autonomous vehicles, and is essential for advancing research in these cutting-edge areas.
“Testbeds determine what we can experiment with, and thereby shape what science questions we can attack,” explained Keahey. “Extending Chameleon from the datacenter to devices that can be combined with IoT peripherals and cost-effectively deployed at the edge has opened up new areas of exploration, especially for projects that use AI in edge to cloud settings; Chameleon makes such projects easy to deploy in non-prescriptive ways, allowing for a wide range of explorations but also for developing and packaging specific solutions thereby accelerating innovation and adoption of new solutions.“
In addition, Chameleon remains dedicated to supporting reproducibility in CS research. The platform will invest in new tools and methodologies to enhance experimental reproducibility, a crucial aspect of scientific research. Ongoing work in systems innovation will ensure that Chameleon continues to provide state-of-the-art experimental capabilities and hardware options.
Finally, Phase 4 will also focus on modernizing and stabilizing Chameleon’s infrastructure. This includes refreshing and upgrading hardware elements and refactoring testbed software and services to align with the evolving software ecosystem. These improvements are essential for managing the accelerating community growth and ensuring that Chameleon remains a leading platform for experimental methodology development.
A PLATFORM FOR FUTURE DISCOVERIES
Chameleon’s role as a deeply reconfigurable and highly utilized edge-to-cloud testbed is vital for CS systems research, education, and emergent applications. The platform’s configuration as a cloud, with reconfigurability at the bare metal level, is critical for performance or power management experiments. Its robust infrastructure, located at UChicago and the Texas Advanced Computing Center at The University of Texas at Austin, connected by high-speed network links, will continue to support a diverse range of research projects.
“The NSF is pleased to continue funding Chameleon,” said Deep Medhi, Program Director for NSF’s Division of Computer and Network Systems. “This is a great resource for the research community to conduct cutting-edge experimental research.”
For more information about Chameleon, please visit www.chameleoncloud.org.