An aggregator combines multiple results produced from one original message. It must receive every related result at the same destination; otherwise, it may fail to merge them or combine unrelated results unnoticed. section 6.3.4
Joining obfuscated and enriched data
A manager splits a message between a data-obfuscation worker and a data-enrichment worker. Both workers send their results to an aggregator. Consistent-hashing routers select the same aggregator using the original message identifier, keeping the two pieces together. section 6.3.4
Trace how parallel worker outputs can reach different aggregators and why message affinity is required. section 6.3.4
Preserve aggregation affinity by:
- Defining a stable identifier for each original message. section 6.3.4
- Routing worker results with consistent hashing so matching identifiers select the same routee. section 6.3.4
- Using a group router when routees register or deregister dynamically. section 6.3.4
