XStream

PERFORMANCE BENCHMARKS

Measure the complete flow.

Real Scala/Python operators, identical workloads, independently verified outputs. Compare file handoffs, model tool scheduling and streaming execution.

Measured on 5 November 2026 · 250,000 observation rows · 48 runs including warmups

End-to-end execution time

Seconds, median of successful measured runs. Lower is faster.

250,000-row workloads
WorkloadAirflow filesLangGraph + model toolsXStream isolatedXStream fused
CSV filtering31.28438.70220.94820.679
Station Join29.01644.49430.44325.890
Grouped aggregation24.09534.07217.30919.616

Failed runs including warmups: 0. Failed trials remain in raw records and are excluded from successful-run medians.

What each method does

Airflow files

Airflow runs a real standalone scheduler and LocalExecutor. Scala/Python containers exchange Arrow IPC files; XCom carries paths. Retries are disabled.

Model tool scheduling

LangGraph asks the model to choose ready steps, executes actual Scala/Python tools and returns result handles. API waits and the final model response are included. Data payloads stay outside the model context.

Model: gpt-6.1-sol

XStream streaming

The same Scala/Python operators exchange streams without intermediate data files. Isolated mode uses separate processes; fused mode places compatible steps in a shared JVM.

Measured figures

Bars show successful measured medians; whiskers span minimum to maximum, not confidence intervals.

Job time
Job time
First data latency
First data latency
Intermediate file volume
Intermediate file volume
Operator processes
Operator processes
Completion counts
Completion counts
Model orchestration breakdown
Model orchestration breakdown

Model overhead and reliability

Measured model runs; successful-run medians
WorkloadSuccessful / attemptedAPI wait (s)Total tokens
CSV filtering3/314.4583710
Station Join3/316.2134193
Grouped aggregation3/314.3223840

API latency depends on the provider, network and load. Token counts are usage, not a price estimate.

Read the results in context

Every method executes the same fixed DAG using the same Scala/Python operator packages and runtime images. This measures runtime scheduling, not the quality of natural-language workflow generation.

Each combination has one warmup and three measured attempts. Scheduling, container startup, computation, serialization and complete output consumption are timed. Input generation, downloads and framework startup are excluded.

These are specific integrations on a shared 16-vCPU host. Airflow does not require files. Different adapters and process layouts contribute to the results; streaming alone is not always faster. Files use OS buffering without fsync, so this is not a durability or physical disk-I/O comparison.

Every successful output is checked against an independent Python implementation for SHA-256, byte length and row count. The raw records include failures, individual timings, model usage, actual tool-call order and artifact hashes.

Reproduce the benchmark →