Skip to main content
Create an account →
Create an account →

Introduction

Quanton is a fast compute engine for Apache Spark, built by Onehouse. It delivers 4x better price-performance for ETL workloads compared to open-source Spark — without rewriting your jobs.

Drop it into your existing Kubernetes cluster via Helm, point your Spark jobs at it, and run. Your code doesn't change.

What Quanton does​

Faster queries, lower cost​

Quanton uses SIMD-vectorized execution, storage-aware query planning, and background indexing to run Spark queries significantly faster than the open-source engine. On TPC-DS (10TB, 99 queries), Quanton completes queries in roughly a quarter of the time — translating to ~73% lower compute cost for the same workload. See Apache Iceberg on Quanton: 3x faster Apache Spark workloads for a detailed breakdown.

Pricing is simple: you pay per GB processed. No per-query fees, no seat licenses, no surprises.

Kubernetes-native​

Quanton is a Kubernetes operator built on the kubeflow SparkApplication CRD. It runs on any conformant cluster — EKS, GKE, AKS, or a local minikube. Your existing SparkApplication manifests and job configs work without modification.

Embedded AI engineer​

An AI assistant runs directly inside the Spark UI with live access to your job's configs, runtime metrics, heap, and GC logs. It's backed by a knowledge server trained on a decade of Spark and lakehouse expertise — so it can diagnose real problems, not just quote documentation. Learn more about Agentic AI →

How it fits into your stack​

Quanton replaces the Spark execution layer. Your data sources, job definitions, orchestration, and catalogs stay exactly where they are. The operator handles scheduling and resource management on Kubernetes; Quanton handles making each job run fast.

Your Spark jobs (unchanged)
        │
        ▼
  Quanton Operator  ←  Kubernetes (EKS / GKE / AKS / minikube)
        │
        ▼
  Your data lake  (S3, GCS, ADLS, or local)

Get started​

Choose a path based on your environment:

PathWhen to use
Local QuickstartTry Quanton on your laptop with minikube — takes ~15 minutes
EKS GuideDeploy on Amazon Elastic Kubernetes Service
GKE GuideDeploy on Google Kubernetes Engine
AKS GuideDeploy on Azure Kubernetes Service
BenchmarksSee TPC-DS results and full performance comparisons

All cloud deployments follow the same pattern: create a cluster, install the Quanton Helm chart using your onehouse-values.yaml, then submit a Spark job. The guides walk through each step.