Bio

Selected projects. Much of the source is visible at github.com, my username is rminnich. Rust for firmware github.com/oreboot/oreboot Firmware is the first thing a CPU runs. Exploits and errors in firmware become exploits and errors everywhere up the stack. Oreboot is a project to explore how we can use a memory-safe language (Rust) for firmware. Oreboot is now in evaluation at several companies. LinuxBoot linuxboot.org LinuxBoot (a.k.a. NERF) restructures firmware to implement separation of privilege at boot time; integration of Silicon Root of Trust (e.g. Open Titan); and better secure the supply chain by removing binary blobs. LinuxBoot is deployed at scale at Google and other companies. VMthreads github.com/akaros/vmm-akaros VMthreads is a new virtualization model in which virtual machines are as easy to start and manage as threads. A virtual machine manager can be written in less than 20 lines of code, and VMs can be started and run for milliseconds, with no special privilege or external daemons needed. We have written proofs-of-concept for Akaros and OSX (the latter in Rust). u-root u-root.org u-root is a set of over 130 standard Unix commands (and growing) written in Go. It is designed for compact root file systems and firmware embed- ding, and is currently deployed in several million datacenter systems. emulytics https://www.academia.edu/16807156/Ten_million_and_one_penguins_or_lessons_learned_from_booting_millions_of_virtual_machines_on_HPC_systems Emulytics uses HPC scale resources to run 10s of thousands to millions of virtual machines to study malware at scale. I started this activity at Sandia in 2007 and it is still in use and producing results. HPC OS https://link.springer.com/article/10.1007/s00450-009-0070-z My team ported Plan 9 to the Blue Gene series of supercomputers. Plan 9 was attractive at the time as it was a much less monolithic kernel than Linux, but arguably more general purpose than IBMs CN kernel. We demonstrated a working port in less than a year of effort. In 2006, we were the first team to port a non-IBM kernel to Blue Gene. Linux HPC https://www.researchgate.net/publication/4087240_Pink_a_1024-node_single-system_image_Linux_cluster My team built the Clustermatic, the first large-scale Linux HPC system at Los Alamos. We showed we could remove all per-node disks; boot in seconds, not tens of minutes; and make upgrading, rebooting, and instal- lation equivalent operations. Clustermatic needed so little maintenance that uptime was >100% as measured by DOE standards.