Audit Log Service · View 33 of 40 · 6 · Operations
The finding
- The requirement's target of $0.09 per million events, all-in for 13 months, implies a total spend of about $5.9 k a month at 66 billion events a month. The segment store alone consumes roughly $4.1 k of that before a single core is switched on, and the cross-region replication the same requirement mandates consumes the rest.
- The defensible figure for this design is about $0.65 per million with a 90-day index and about $0.75 with a full 13-month index — roughly seven to eight times the target.
- Recommendation: renegotiate the target rather than the architecture. Every property that could be dropped to approach $0.09 — DR, HSM-backed signing, the separately administered ledger — is a property the requirement asks for by name elsewhere.
The arithmetic
- Capture $12.5 k · Custody $14.7 k · Index $5.5 k (90 d) · Serve and prove $6.0 k ≈ $42.7 k a month, over 66,000 million events.
- A fixed floor of about $15.6 k a month — Managed HSM in two regions, the Confidential Ledger, PostgreSQL, observability and a minimum broker and cluster footprint — does not scale down with volume. At this volume that floor alone is $0.24 per million.
- The index is the dominant marginal cost, which is why its window is tunable per tenant independently of retention. Moving the default from 13 months to 90 days saves about $6.5 k a month for a query-path change most tenants will never notice.
Levers, in order of value
- Per-tenant index window, defaulting to 90 days (largest single saving).
- Chain-preserving compaction of small segments, which cuts both storage transactions and the small-object tail.
- Charging query cost to the querying tenant with a per-role budget, which stops the unbounded 13-month scan being someone else's problem.
- Archive tiering from 13 months, worth roughly $2 per TB-month against $10 in cool storage.