load-balancing

Distributing a Request Batch Across Instances with FORK_JOIN

A batch of requests arrives and needs to be processed across multiple service instances in parallel to reduce total latency. This workflow uses FORK_JOIN to distribute partitions to 3 instances, each processing its share on a different host, then aggregates the results with total processed count and average latency.

Workflow

requestBatch
     |
     v
  FORK_JOIN: lb_call_instance (3 instances)
    instance-1: partition A on host-1
    instance-2: partition B on host-2
    instance-3: partition C on host-3
     |
     v
  JOIN
     |
     v
+-----------------------+
| lb_aggregate_results  |   totalProcessed, avgLatency
+-----------------------+

Workers

CallInstanceWorker -- Processes a partition on a specific host. Returns a result map with the instanceId, partition, and processing details.

AggregateResultsWorker -- Combines results from all instances. Returns totalProcessed, avgLatency, and the aggregated result list.

Tests

16 unit tests cover instance calls and result aggregation.

Running

See ../../RUNNING.md for setup and execution instructions.