Real-Time vs micro-batch inside WebAssembly

One persistent Spark WASM engine. Two isolated scheduling arms. Identical deterministic telemetry and Spark SQL, entirely inside this browser tab.

Arm duration (ms)

Input rows / second

Resident source tick (ms)

Micro-batch trigger (ms)

Loading Spark WASM…The first engine compile may take several seconds.

Resident-style Real-Time

Waiting…

Processing-time micro-batch

Waiting…

Event-generation → Spark SQL completion (p95)

Real-Time

—

Micro-batch

—

Resident-style arm

100 ms source tick→immediate Spark SQL→completion receipt

No trigger wait after a row becomes available.

Micro-batch arm

100 ms source tick→trigger buffer→Spark SQL batch

Rows wait for the configured processing-time trigger.

Parity-checked final Spark SQL result

category	events	avg value	peak value

Comparison events

Browser-safe semantics demo: this compares immediate resident-style scheduling with processing-time buffering over the same WASM Spark SQL engine. It does not embed native Kafka, librdkafka, sockets, or the native spark-streaming runtime in the browser.