Compare Scenarios

Migration Step 1

Both scenarios define a migration step at this stage. Financial Ledger Platform: triggered by 'Audit requirement to reconstruct account state at any histor'. ML Feature Serving Platform: triggered by 'Feature computation logic duplicated across 3+ model serving'.

Migration Step 2

Both scenarios define a migration step at this stage. Financial Ledger Platform: triggered by 'Kafka publish failures causing financial transaction rollbac'. ML Feature Serving Platform: triggered by 'Regulatory or reproducibility requirement to re-run model pr'.

Migration Step 3

Both scenarios define a migration step at this stage. Financial Ledger Platform: triggered by 'Financial dashboard query p99 > 500ms causing dashboard-driv'. ML Feature Serving Platform: triggered by 'Semantic similarity search recall insufficient from BM25 ful'.

pg_locks shows contended rows on accounts table; write p99 > 50ms; deadlock errors in application logs; pg_stat_activity showing many transactions waiting for RowExclusiveLock on the same account rows

Tier 1: Hot Account Lock Contention: Concurrent debit/credit transactions competing for the same account row versions. Recommended evolution: Implement optimistic locking with version column and retry; or queue concurrent updates for the same account entity through an account-scoped serialization queue at the application layer; or partition the accounts table by account range .

Write p99 > 100ms with synchronous_commit = remote_apply; replica WAL apply lag visible in pg_stat_replication; network jitter between primary and replica causing write latency spikes correlating with replication ACK delays

Tier 2: Synchronous Replication Write Latency: Synchronous replication write-ahead wait amplifying network latency for every committed transaction. Recommended evolution: Co-locate primary and replica in the same availability zone for lowest replication RTT; tune wal_sender_timeout and recovery_min_apply_delay; evaluate whether synchronous_commit = on (durable to primary WAL only) is acceptable for your regulatory risk model .

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. .