Supercomputer RAMSES Enables End‑to‑End Encryption of Data‑in‑Use for Genomics Workloads
What Happened — Researchers at the University of Cologne unveiled RAMSES, a prototype supercomputer that keeps data encrypted even while it is being processed. The system stitches together AMD’s hardware memory‑encryption (SEV), IBM’s storage‑encryption software, Thales’s key‑management appliance, and Cisco Duo MFA. In two genomics benchmark jobs the strongest security posture added only ~4.4 % latency.
Why It Matters for Compliance & Audit Readiness
- Provides a concrete method to satisfy SOC 2 CC6 (Confidentiality) and CC7 (Privacy) controls that require protection of data in use, a gap that many traditional audit programs overlook.
- Because encryption is performed by the processor and keys are managed in a dedicated HSM, the control can be continuously monitored and automatically captured as audit evidence, reducing reliance on manual checks.
- Aligns with Verisq’s Control Mapping capability: you can map the hardware‑based memory‑encryption control to your SOC 2 control matrix and generate continuous proof for auditors.
Who Is Affected — Enterprises that run high‑performance compute (HPC) or cloud‑based analytics on sensitive data, especially in health‑life, biotech, academic research, and any sector processing regulated information at scale.
Recommended Actions
- Assess your current data‑in‑use protection gap against SOC 2 CC6/CC7 requirements.
- Pilot confidential‑computing hardware (e.g., AMD SEV) and integrate it with a proven key‑management solution (Thales) and MFA (Cisco Duo) in your job‑submission workflow.
- Document the configuration and collect automated logs/evidence to satisfy continuous‑compliance and audit‑readiness checks.
Technical Notes — RAMSES leverages AMD’s Secure Encrypted Virtualization (SEV) for transparent memory encryption, IBM’s storage‑encryption stack, Thales’s HSM for key storage, and Cisco Duo for second‑factor authentication. Performance impact measured at a 4.4 % slowdown on I/O‑heavy genomics workloads. Source: Help Net Security