Legacy Cloud Infrastructure and Reprogrammable FPGAs
April 21, 2022

Legacy infrastructure contains years of business meaning inside variable names, tables, classes, procedures, configuration, and operational habits. A useful migration platform must preserve that meaning while reorganizing execution for massive parallelism.

Brian Greenforest proposes a self-reprogramming FPGA cloud that turns the entire backend into one elastic, reactive, object-oriented hardware system.

Preserve Meaning While Changing the Machine

A migration compiler can map arbitrary languages, databases, operating-system configuration, and service relationships into a common mathematical model. It can retain familiar names and data structures so engineering teams still recognize the system they own.

The runtime then places operations, data, and communication across reconfigurable regions. Explicit object capabilities govern access and ownership while reactive connections carry changes through the system.

Put Services Beside the Work That Uses Them

A shared supercomputer chip can host customer backends and low-latency service APIs on the same fabric. Local interaction replaces repeated trips across congested networks and makes service-oriented architecture physically immediate.

Organizations with valuable legacy estates can bring a representative workflow into this architecture. The migration path starts with preserving semantics and ends with a system that can reshape itself around demand.

Execute Legacy Cloud Infrastructure and Reprogrammable FPGAs Through Cartilage Runtime Instantiation

Cartilage Runtime-Instantiation Architecture carries this system into nested modules that enter and connect inside a live spatial machine.

Cartilage Runtime-Instantiation Architecture

Originally posted on LinkedIn

Brian Greenforest · (2022-04-21 17:50:53 UTC)

Open the original LinkedIn post · LinkedIn activity 6922964896301162496

LinkedIn status when archived: Visible to anyone on or off LinkedIn.

You can't port your legacy cloud infrastructure easily to any other platform, unless it's a generalized superscalable supercompute cluster of self-reprogrammable FPGAs, because only that model allows to reason about ANY ARBITRARY programming language, database structure, and even the OS config, in a generic, mathematical way, doing automatic parallel and concurrent reorganization of your legacy ported code, while preserving old variable names, function names, class names, table names, etc. The result is dynamic scalable reactive object-oriented, PERFECTLY SECURE, elastic to scale DIGITAL CIRCUITS that run your #database directly in #silicon in the cloud, in a versatile elastic pay-for-what-you-use way, while accessing unprecedented low-latency proprietary cloud APIs from major Fortune-500 companies! SOA become real on a supercomputer chip. Instead of wasting energy on traversing the already busy internet, you can consume the APIs of major vendors directly on the same chip, your backend is running at!