Integration Ecosystem

CLIO Plugins

30+ built-in integrations. Pipeline-based deployment. Turn complex HPC applications into composable, reproducible workflows in minutes.

How It Works

CLIO Plugins are powered by Jarvis-CD — a unified deployment framework that turns any application into a composable pipeline package.

1

Discover Resources

Auto-detect storage devices, network topology, and hardware capabilities across your cluster.

jarvis rg build
2

Compose Pipelines

Chain plugins into multi-stage pipelines. Storage + simulation + analysis + visualization — all in one workflow.

jarvis ppl append lammps
3

Deploy Anywhere

Same pipeline definition works on a laptop, a Slurm cluster, or a containerized cloud environment.

jarvis ppl run

Plugin Catalog

Every integration is a self-contained package with automated configuration, dependency management, and multi-node support.

🔬 Scientific Simulations

LAMMPS

Molecular dynamics — materials, polymers, biological systems

Gray-Scott

Reaction-diffusion simulation via ADIOS2

Gadget-2

N-body/SPH cosmological simulations

CM1

Cloud-resolving atmospheric model

InCompact3D

High-order incompressible flow solver

Nyx

AMReX-based Lyman-alpha cosmology

Cosmic Tagger

Deep learning for neutrino physics

DDMD

Deep-learning driven molecular dynamics

PyFlexTrkr

Feature tracking for climate science

💾 Storage Systems & I/O

OrangeFS

Parallel file system deployment and management

Redis

In-memory data store for caching and messaging

Spark Cluster

Apache Spark for distributed data processing

Data Stage-In

Automated data staging for pipelines

MkFS

Filesystem provisioning and formatting

📊 Benchmarks & Analysis

IOR

Parallel I/O benchmarking (POSIX, MPI-IO, HDF5)

FIO

Flexible I/O tester for storage performance

FileBench

Workload model-driven filesystem benchmarks

DLIO Benchmark

Deep learning I/O characterization

Darshan

I/O tracing and performance analysis

ParaView

3D scientific visualization pipelines

Compose Real Pipelines

Chain plugins to build end-to-end scientific workflows. Here's a molecular dynamics pipeline with tiered storage and live visualization:

# Initialize Jarvis for your cluster

jarvis init ~/jarvis_conf ~/jarvis_local ~/jarvis_shared

# Discover available storage (NVMe, RAM, PFS)

jarvis rg build

# Build the pipeline: IOWarp runtime → LAMMPS → Darshan → ParaView

jarvis ppl create md_analysis

jarvis ppl append iowarp_runtime

jarvis ppl append lammps

jarvis ppl append darshan

jarvis ppl append paraview

# Deploy across 4 nodes and run

jarvis hostfile set ~/cluster_hosts

jarvis ppl run

Plugin Architecture

Resource Graph

Automatic hardware discovery. Plugins query the resource graph to find available storage devices, network interfaces, GPUs, and compute nodes.

No manual configuration for different machines — Jarvis adapts to what's available.

Pipeline Composition

Plugins chain into multi-stage pipelines with dependency resolution. Each package manages its own lifecycle: configure, build, start, stop, destroy.

Shared/private directories enable multi-node data sharing without manual orchestration.

Interceptors

Transparently intercept I/O calls (POSIX, MPI-IO) to route through IOWarp's storage engine — no application changes required.

Legacy applications get IOWarp benefits with zero code modifications.

Custom Plugins

Build your own plugin in Python. Extend the JarvisPackage base class, define configuration, and plug into any pipeline.

Repository system for sharing plugins across teams and organizations.

Build Your Own Plugin

Every scientific application can become a CLIO Plugin. Start with our example template, define your configuration schema, and your app is ready to compose into any IOWarp pipeline.