Kubernetes vs Nomad vs Expanso – Choose the Right Orchestrator

Kubernetes vs Nomad vs Bacalhau: Which Orchestrator is Right for Your Data?

If you are an engineer trying to manage petabytes of data scattered across sources, you know the pain of dragging it all to a central cluster. It takes days—if not weeks—and costs a fortune in bandwidth. Traditional orchestrators like Kubernetes and Nomad assume you will move data to compute. But data gravity, sovereignty, and edge scale make that slow, costly, and risky.

Instead of moving data to compute, Bacalhau brings compute to the data—whether it’s on edge devices, in the cloud, or in your own servers. The result: faster processing, lower costs, and improved security.

This guide covers:

• What Kubernetes, Nomad, and Bacalhau are
• The key differences among them
• When to choose each
• Why Bacalhau might be the missing piece in your infrastructure

Kubernetes, Nomad, Bacalhau: Core Concepts and Goals

Kubernetes

Kubernetes is the industry-standard container orchestrator, offered as a hosted platform by major clouds. It shines with rich ecosystems like Helm, Istio, and Prometheus, and is built for microservices that mesh internally.

Nomad

Nomad is lightweight and flexible. It runs containers, VMs, and standalone binaries. Because it lacks an internal service mesh, it needs fewer supporting services, which simplifies setup across AWS, GCP, Azure, and on-prem.

Bacalhau

Bacalhau extends the lessons of Kubernetes and Nomad to today’s global, data-heavy layouts. You can run compute next to your data—whether it’s fresh off a sensor, in an S3 bucket, SQLite, Iceberg, or streaming from edge devices. It reuses existing hardware, reduces data transfers, and supports Docker, WASM, and custom binaries without forcing app rewrites.

Comparing Kubernetes, Nomad, and Bacalhau: Key Differences

Execution Environments

• Kubernetes: Containers first (Docker and other runtimes); WASM emerging; diverse runtimes often need extra tooling.
• Nomad: Containers plus non-containerized apps (VMs, static binaries, JARs, more) via drivers.
• Bacalhau: Native Docker and WASM; pluggable executors for custom binaries and hardware (including GPUs).

Lightweight Execution & Mixed Architectures

• Kubernetes: Powerful but control-plane heavy; mixed arch requires careful config.
• Nomad: Single binary for server/client; constraints let you target architectures.
• Bacalhau: Agents run nearly anywhere; schedule jobs to the nodes you provide, regardless of runtime.

Job Types

• Kubernetes: Services (Deployments/StatefulSets) and batch (Jobs/CronJobs). Large batch often delegated to Spark/Flink atop K8s.
• Nomad: Service, batch, and system schedulers.
• Bacalhau: Batch, service, ops, and daemon jobs; embarrassingly parallel by default for throughput.

Queued Execution

• Kubernetes: Scheduler queues when resources are scarce; higher-level batch systems add their own queuing.
• Nomad: Pending jobs enter evaluation queues until capacity matches constraints.
• Bacalhau: Explicit job queuing with configurable backoff and timeout when matching nodes are busy or disconnected.

Disconnected Execution

• Kubernetes: Handles pod/node failures; edge/disconnected needs integrations.
• Nomad: Clients can continue running during partitions with a disconnect block; oriented to central control.
• Bacalhau: Built for intermittent connectivity and edge; jobs can queue or fail fast based on policy.

Security

• Kubernetes & Nomad: Strong controls (RBAC/ACLs, network policies, secrets) but centralized data/metadata increase exposure.
• Bacalhau: Compute-over-data minimizes movement; metadata isn’t centralized by default, reducing leakage risk.

Quick Comparison

Choosing Your Orchestrator: Kubernetes vs Nomad vs Bacalhau

When to Choose Kubernetes

Industry standard for complex, containerized microservices at scale with rich networking, service discovery, autoscaling, and self-healing. Great for tightly coupled services in a single region or cloud.

When to Choose Nomad

Choose Nomad for simplicity and flexibility inside a single zone or region when Kubernetes feels heavy. Ideal if you need one orchestrator for containers, VMs, and binaries without the full K8s control-plane overhead.

When to Choose Bacalhau

Bacalhau specializes in Compute Over Data for distributed or edge datasets:

• Distributed datasets: Avoid hauling petabytes to a central cluster; run compute where data lives (e.g., satellite imagery across regions).
• Data gravity/sovereignty: Keep data in-place for regulatory or policy reasons (e.g., hospital data across jurisdictions).
• Edge processing: Run analytics/ML inference at the source (e.g., industrial sensors for real-time anomaly detection).

Conclusion

There’s no single “best” orchestrator—only the best fit for your workload. Kubernetes dominates cloud-native microservices. Nomad offers a lean, flexible scheduler for mixed workloads without heavy control planes. Bacalhau excels when data movement is the bottleneck, embracing Compute Over Data to cut cost, latency, and risk while enabling WASM sandboxing and hardware-aware execution.

What’s Next?

To try Bacalhau, install it and run your first jobs. If you don’t have a network ready, start with Expanso Cloud, or set up your own cluster with the network setup guides.

Get Involved!

• Expanso Website
• Bacalhau Website
• Bacalhau on Bluesky
• Bacalhau on X
• Expanso on X
• TikTok
• YouTube
• Slack
• LinkedIn
• Careers

Commercial Support

Bacalhau is open source, with binaries built, signed, and supported by Expanso. For commercial support or pre-built binaries, contact us or get your license on Expanso Cloud.