A configurable experimental environment for large-scale edge to cloud research

Traditional distributed databases are often slowed down by network communication overhead. The Tigon project introduces a new database design that tackles this bottleneck using Compute Express Link (CXL), a technology that allows multiple computer hosts to access a shared memory pool. Tigon employs a hybrid approach, keeping most data in fast, local host memory while moving only actively shared data to the CXL memory. This results in significant performance gains, achieving up to 2.5 times higher throughput than traditional databases. Since the multi-host CXL hardware required for this research was not yet commercially available, the project was brought to life by emulating the next-generation environment on Chameleon's bare-metal servers.

Have you completed the Chameleon Getting Started guide and are wondering what comes next? This post will help you dive deeper by moving beyond the basics. You will learn how to interact with the Chameleon API from the command line on an instance and from your laptop. We'll also cover how to use tools like rclone and cc-snapshot to securely save and manage your valuable research data.

This month, we have a significant change to KVM@TACC: as per our announcement last month, all KVM instance launches will now require an advance reservation step in the allocation workflow (with the “advance” potentially being once second from now, i.e., on demand). Also, we have cool new hardware on CHI@Edge, and improvements to the “message of the day” on Chameleon-supported base images