It’s easy to envision on premises High Performance Computing (HPC) budget and strategy meetings focusing on compute and storage bandwidth, rather than consider complex infrastructure overhauls. “We need to increase compute cycle efficiency.” An easy to understand and quantifiable goal met with nods of approval, HPC efficiency is the paramount concern. Working alongside compute performance considerations, CISOs must consider responses that fit to more esoteric challenges — minimizing cyberthreat liability, an increasingly obscure and insidious adversarial threat. This is no small challenge, and the risk in failure is quantified directly by a multi-billion dollar Ransomware-as-a-Service (RaaS) industry.
The responsibility of protecting massive proprietary stores of data for analysis, when most thoughtfully considered, begins with prioritizing confidentiality, integrity, and availability (CIA). Critically valuable supercomputing systems are often air gapped, requiring a hand-built configuration of infrastructure to be implemented on premise. Unfortunately, tweaks and tunes must be manually configured, and mistakes could be made during that process.
Public cloud allows any business to have IT without any technical prowess, they already have a big shiny virtualized automated 3D printer ready to spit out an IT asset of any size and scale. The dichotomy for businesses is in the imbalance of flexibility and security, having to choose one or the other begs for a market correction. HPC solutions in the cloud automate infrastructure for flexibility and on premise solutions air gap for security. Lawyers and security auditors love complete isolation, too, but the challenges of manual configuration within an air gap are substantial — how can the lawyers and the IT staff both get a win?
The problem for many is there is no clear path to cutting the cord. For the scissor-less, there are a few major competitive disadvantages they must accept using CSP infrastructure; lock-in from sticky dev environments and tool kits, “data gravity” issues that make ingress/egress expensive, and the risk of breaches and outages. However, if those two circles could meet in a platform to deploy autonomous clouds on bare-metal anywhere, it would be the best of both worlds. AI/ML/Data Science systems are most economical and secure in an air gap. Put down the chisel and flip the switch on the 3D printer. Bridging the gap, the NetThunder platform is an on premise private cloud with the automation needed to make cutting the cord practical — why whittle out a secure network when it can be 3D printed?
Any well implemented HPC project, manufacturing system, sensitive data lake, etc., will have a holistic balance of infrastructural security and flexibility at its foundation. A truly flexible infrastructure is able to deploy and operate anywhere, even after a pair of scissors cuts it off from the world.