Compare Scenarios

Migration Step 1

Both scenarios define a migration step at this stage. ML Feature Serving Platform: triggered by 'Feature computation logic duplicated across 3+ model serving'. Search-Heavy Content Platform: triggered by 'Search query p99 > 500ms on full-text queries; faceted navig'.

Migration Step 2

Both scenarios define a migration step at this stage. ML Feature Serving Platform: triggered by 'Regulatory or reproducibility requirement to re-run model pr'. Search-Heavy Content Platform: triggered by 'Application write latency increasing due to Elasticsearch in'.

Migration Step 3

Both scenarios define a migration step at this stage. ML Feature Serving Platform: triggered by 'Semantic similarity search recall insufficient from BM25 ful'. Search-Heavy Content Platform: triggered by 'High write throughput (> 10k documents/min) causing segment '.

Feature store p99 rising from < 5ms to > 50ms immediately following a model deployment or pod scaling event; Cassandra or PostgreSQL feature store read QPS spiking 10–100x simultaneously with serving fleet restart; feature store connection pool exhaustion visible in application logs during the cold start window; Redis cache hit rate dropping to < 10% for the first 60 seconds after deployment

Tier 1: Cache Cold Start Latency Spike: Simultaneous cache cold start across all serving pods after deployment, converting sequential feature requests into a thundering herd against the backing feature store. Recommended evolution: Implement pre-warming: before a new model version receives traffic, a warm-up job runs inference requests for a representative sample of entity IDs, populating the Redis cache before the pod enters the serving fleet. Use probabilistic early cache refresh (fetch from backing store before TTL expiry when remaining TTL < 20% under high request rate) to prevent simultaneous TTL expiry on hot features. Stagger deployment rollout: deploy 10% of pods, wait for cache warm, then proceed to the next 10%. .

Model performance metrics (precision, recall, AUC) degrading without a corresponding input distribution shift; ClickHouse drift dashboard showing feature value distributions served online diverging from training population distributions; explicit skew audit (comparing offline training feature values against replayed online feature values for the same entity at the same timestamp) showing systematic differences in specific features

Tier 2: Training-Serving Skew Detection: Feature computation logic divergence between the offline training pipeline and the online serving pipeline: a feature transformation applied in training is not applied identically in serving. Recommended evolution: Enforce a single feature computation function registered per feature in a shared feature registry, executed by both the online pipeline and the offline training pipeline. The two paths must use the same code, not independently maintained implementations. Implement a skew monitoring job that computes a random sample of features using both paths and alerts on distribution divergence above a threshold. .

Elasticsearch index CDC consumer lag > 10s; search results showing items that no longer exist or missing recently published items; CDC connector health dashboard showing processing rate below write rate

Tier 1: Index Freshness Degradation: CDC consumer or Elasticsearch bulk indexer not keeping pace with PostgreSQL write rate. Recommended evolution: Increase Elasticsearch bulk indexer thread count; tune bulk index batch size and flush interval; profile CDC connector bottleneck (network vs Elasticsearch write throughput vs mapping complexity) .

Elasticsearch JVM heap usage > 75% sustained; GC pause events visible in cluster logs; query p99 latency spikes during GC; cluster health showing yellow (unassigned shards during GC recovery)

Tier 2: Search Cluster Heap Pressure: Large aggregation queries or high document count per shard exceeding JVM heap budget. Recommended evolution: Increase Elasticsearch heap to 50% of node RAM (max 30GB for ZGC); reduce shard count to keep per-shard document count < 50M; disable dynamic mapping and explicitly define all field types; move to doc values for all non-analyzed fields .