Infrastructure consulting for research computing, cloud and AI.
Fenn Enterprises is the independent practice of Michael Fenn, a systems engineer whose work spans HPC clusters, virtualization, high-performance storage and networking, and cloud cost and performance analysis. Clemson alumnus. Co-author on the Anton 3 paper named a Gordon Bell Prize finalist at SC21.
What I help with
Cloud and AI infrastructure
Designing and right-sizing compute for simulation, data and model workloads on AWS and Kubernetes — including honest cost-versus-performance analysis before you commit, in the spirit of the EC2 cluster compute study below.
HPC clusters and workload management
Cluster architecture, job scheduling and virtual clusters that stretch across sites. Long experience with grid and cloud provisioning, from Virtual Organization Clusters to modern schedulers.
Storage and network performance
Evaluating and tuning the fabric your jobs actually run on: iWARP versus InfiniBand, RDMA best practices, and caching network filesystems such as fs123 for large read-mostly datasets.
Virtualization and Linux platforms
KVM and vSphere environments, secure remote administration, and configuration management with Bcfg2 and Ansible so systems are reproducible rather than hand-built.
Track record
- Gordon Bell Prize finalist, SC21, as a co-author of “Anton 3: Twenty Microseconds of Molecular Dynamics Simulation Before Lunch.”
- Senior Research Personnel on a $497,178 National Science Foundation CC-NIE award to build out a dedicated research and education network for big-data science.
- Invited speaker at USENIX LISA 19, the Ethernet Alliance webinar series, the OpenFabrics Alliance International Workshop, and a Supercomputing 2010 panel on computational science education.
- Eleven peer-reviewed publications on cloud, grid and cluster computing, plus industry white papers on interconnect performance and public-cloud economics.
- Systems administration at D. E. Shaw Research, supporting one of the most demanding molecular-dynamics computing environments in the world.
Selected publications and talks
- 2021Anton 3: Twenty Microseconds of Molecular Dynamics Simulation Before Lunch SC21, ACM — Gordon Bell Prize finalist
- 2019fs123: An Open-Source, Network Filesystem with Pervasive Caching Invited talk, USENIX LISA 19
- 2014Pulley: Secure Administration of Virtual and Remote Computing Systems ACM Southeast Conference
- 2012iWARP Learnings and Best Practices Invited talk, OpenFabrics Alliance International Workshop
- 2011Evaluation of iWARP versus InfiniBand Performance White paper
- 2010A Performance and Cost Analysis of the Amazon EC2 Cluster Compute Instance White paper
- 2010Scaling Virtual Organization Clusters over a Wide Area Network using the Kestrel Workload Management System VTDC 2010
- 2010Autonomic Clouds on the Grid Journal of Grid Computing
- 2009Dynamic Provisioning of Virtual Organization Clusters IEEE CCGrid ’09
- 2008An Evaluation of KVM for Use in Cloud Computing ICVCI ’08
The full list, with links to papers and slides, is on mfenn.com.
Open source
- fs123 — a network filesystem with pervasive caching, built for read-mostly scientific data.
- Bcfg2 — configuration management for large Linux fleets.
- Ansible vSphere module — automation for VMware environments.
Practical write-ups on IPMI, vSphere, Open Science Grid and disk-image conversion are collected as HOWTOs.
Working together
Engagements range from a short architecture review to hands-on build-out of a cluster, storage tier or cloud environment. The first conversation is free and usually enough to tell whether it's a fit.
Email michael@fenn.cc, or start with the résumé for a fuller history.
Fenn Enterprises7151 O'Kelly Chapel Rd #212
Cary, NC 27519