Colloquium Archive

Assumptions: They Make An… Insecure System

Jason Lowe-Power, UC Davis


In this talk, I will describe how Meltdown and Spectre break programmer expectations and cause a security nightmare. Meltdown and Spectre shocked the IT world when first publicly reported 9 months ago because these vulnerabilities were not bugs or errors in the hardware. Instead, these vulnerabilities target the performance optimizations used in high performance processors for the past 25 years. In this talk, we'll look at how these performance optimizations break programmers' assumptions of how their programs execute and can lead to confidential data leaking in mobile devices, personal computing devices, and shared cloud platforms.

High-Performance Computing and Livermore Computing Division

Rigoberto Moreno Delgado and Jean Shuler
Lawrence Livermore National Laboratory


The Livermore Computing Division (LCD) mission is to develop and support the Lawrence Livermore Laboratory high performance scientific computing environment, including world-class supercomputers, file systems, archival storage, data analysis resources, and other production-quality, secure services. LCD develops tools, system software, and an application infrastructure that enables effective use of computing resources. The LCD staff provide user services, training, documentation, and consulting that support usability, accessibility, and reliability for the computing systems.

The HPC Cluster Engineer Academy is a 9-week paid internship that will give you direct experience with running and maintaining high performance computing (HPC) systems. Although it doesn’t get much attention in traditional academic coursework, cluster engineering is a main focus of HPC centers around the world. For every software developer or computational scientist who enjoys a fast, reliable, and up-to-date computer system, there’s a cluster engineer who maintains it. As an HPC Academy intern, you’ll learn the basics of cluster engineering and system installation under the guidance of our experts. This internship provides practical, real-world application of knowledge no HPC center—like LLNL—can live without. This 9-week internship will start at a low level with the basics of shell scripting, distributed version control software, and compiling/porting files on Linux systems. Then we will move on to system installation and eventually we'll get into some networking, security, file systems, and various High Performance Computing topics. The first 5-6 weeks will be spent with lectures combined with testing and hands-on work building a test cluster. The final three to four weeks will be group projects making use of these skills. For the group projects, the students will be divided into teams of students. At the end of the program the students will have a tangible project they present to the Livermore Computing staff.

Survey of Semiconductor Test and Quality

Matthias Kamm


As Moore’s Law continues to progress, it ensures finer geometries, billion transistor designs and increasing complexity. To ensure SoC devices meet customer expectations for quality and reliability a thorough test strategy is required. Companies designing and manufacturing semiconductor SoCs must optimize requirements in different areas such as test time, coverage, yield and cost. Baseline requirements can vary based on the target markets, but general expectations for quality are increasing with every generation of computing device. Topics include semiconductor design, test, quality, reliability, and future technologies being researched by the industry.