Research
Softadastra studies whether software can become less dependent on the machines, networks and infrastructure that execute it.
The work combines programming models, runtimes, distributed systems, state, reliability and heterogeneous computing.
Runtime research implementation ↗Can software be separated from the infrastructure that executes it?
Software should describe what it needs to accomplish, not the infrastructure required to accomplish it.
The research question is not whether infrastructure disappears. It is how much infrastructure knowledge can move out of application code and into the computing system.
Research program
These questions are not presented as solved problems. They define the technical territory that guides the Runtime and related systems work.
Application code still contains many assumptions about machines, regions, databases, accelerators, networks and deployment topology.
Explore programming models where intent, state and constraints can remain separate from infrastructure selection.
Placement, scheduling, recovery and resource selection are often decided before execution begins.
Study a runtime capable of making and revisiting those decisions from application requirements and current system conditions.
Moving computation is easier than moving durable, consistent and locally useful application state.
Explore state models that support replication, recovery, reconciliation and changing execution locations.
Many distributed systems treat connectivity as a permanent assumption even when real environments cannot provide it.
Investigate local-first execution where eligible workloads and state remain available during network partitions.
Applications increasingly depend on specific CPUs, GPUs, accelerators and provider-specific execution environments.
Explore capability-driven resource selection where applications describe requirements rather than permanently selecting hardware.
Complete infrastructure independence may not be possible or desirable for every workload.
Measure which decisions can move safely into the runtime and which constraints must remain explicit in application code.
Research territory
Infrastructure independence cannot be implemented by a single abstraction. The Runtime sits at the intersection of several technical areas.
How software expresses intent, state, capabilities and constraints without encoding unnecessary infrastructure decisions.
How execution systems observe conditions, make decisions and reconsider those decisions while software is running.
How application state can remain durable, available and reconcilable across machines, failures and network partitions.
How computation can be mapped onto available resources using locality, latency, trust, cost and capability constraints.
How systems recover from lost responses, failed processes, unavailable machines and partially completed operations.
How useful computation and state can remain available when remote infrastructure cannot be reached.
How CPUs, GPUs and future accelerators can be selected as capabilities without becoming permanent application dependencies.
How identity, trust and policy constrain where computation and state are permitted to exist.
Softadastra's research is intended to produce executable systems, not only architecture diagrams.
An idea becomes more useful when it can be implemented, benchmarked, broken under failure, observed under constrained resources and compared against simpler alternatives.
That is why open-source implementation work is part of the research process itself.
Research artifacts
We do not currently present a catalogue of academic publications. The primary public evidence of the work is the software being implemented and maintained.
Implementation of the adaptive execution layer at the center of the research direction.
source ↗ maintained Vix.cppOpen-source C++ developer infrastructure used to build and test real systems.
vixcpp.com ↗ public Open-source systemsAdditional build, reliability and runtime systems maintained in the open.
explore →Research position
Infrastructure-independent computing is a direction we are trying to make increasingly practical. Some infrastructure constraints may always need to remain explicit. Some runtime decisions may be too expensive, unsafe or workload-specific to automate.
The purpose of the research is to discover that boundary through working systems.
Work on the problem
If you work on programming languages, runtimes, distributed systems, reliability, state, operating systems, networking or heterogeneous computing, we are interested in hearing from you.