AGUASTO

Distributed inference coordination

NVIDIA Dynamo

Coordinate inference engines through routing, separate prefill and decode workers, cache management and scaling.

Original Aguasto conceptual illustration. Not manufacturer geometry, an operating Aguasto asset, or a measured benchmark.

When it is relevant

  • Coordinate inference across multiple GPUs or nodes
  • Evaluate KV-aware routing or independent prefill and decode scaling
NVIDIA Dynamo official documentation (external source) · Dynamo official repository overview (external source)

Hardware and software assumptions

Documented ecosystem: NVIDIA, AMD, Intel. Platform support is not a guarantee that every checkpoint, optimization or feature works on every device.

  • Use a documented backend and feature combination for the selected release
  • Plan network, storage, cache-transfer and operational requirements
  • This architectural reference is not an Aguasto operating cluster
NVIDIA Dynamo official documentation (external source) · Dynamo official repository overview (external source)

Operating trade-offs

  • Extra coordination layers add deployment and operational complexity
  • An engine alone may be sufficient for a single model on one GPU
  • Cross-vendor support is backend- and feature-dependent
NVIDIA Dynamo official documentation (external source) · Dynamo official repository overview (external source)

Continue the research

Choosing an inference frameworkPrefill, decoding and disaggregationBenchmark methodologyModel referencesInfrastructure collaboration

Sources & verification

Reference revision 1 · Reviewed . Manufacturer claims and documentation are attributed; Aguasto has not run the cited products in production.

  1. NVIDIA Dynamo official documentation ↗

    vLLM, SGLang and TensorRT-LLM integration; NVIDIA / AMD GPU and Intel XPU support statements; disaggregated serving, routing, cache management and scaling. Publication: date not stated in the reviewed source. Verified: 2026-10-12.

  2. Dynamo official repository overview ↗

    Coordination above inference engines; multi-node relevance; single-GPU engine-only guidance; feature support differs by backend. Publication: date not stated in the reviewed source. Verified: 2026-10-12.