How It Works
How it works
Keep your Spark code.
Change the runtime.
Using Spark Connect, Fusion determines the best execution path for each workload. Compatible workloads execute natively in Fusion, while workloads requiring Spark continue through your existing Spark runtime.
YOUR CLOUD · DATA STAYS IN S3, ICEBERG, UNITY
YOUR SPARK CODEAPPS · PIPELINES
SPARK CONNECT01 · DECIDE PATH
NATIVE FUSION EXECUTION02 · COMPATIBLE WORKLOADS
EXISTING SPARK RUNTIME03 · AUTOMATIC FALLBACK
01
Intelligent Runtime Decisioning
Automatically determine the best execution path for each workload.
02
Native Fusion Execution
Run demanding workloads through Fusion's purpose built execution engine using significantly less compute.
03
Automatic Spark Fallback
Maintain compatibility with your existing Spark environment when workloads require it.
Zero data movementNo code portingFewer out of memory failures and retries
Fusion is available as a standalone engine and also powers large scale data processing across the Aqfer Marketing Data Engine.
Interoperability
Works With the Environment You Already Have
Fusion is designed to fit into your existing cloud and data infrastructure without requiring you to replatform.
AWS
Spark workloads on Amazon EKS and Amazon EMR.
Databricks
Existing Databricks and Spark based workloads.
Open data
S3, S3 compatible storage, Iceberg, and Unity catalogs.
Google CloudQ4
GKE and Google Cloud Dataproc.
