Architecture Overview

Traffic and Truth are Different Data Products

Quant Edge Exchange demonstrates how high-volume event traffic and financial truth can be modeled using purpose-built data stores.

Launch Simulator View Analytics

Why Two Databases?

Traditional Architecture

Every event enters the same database regardless of whether it is temporary traffic or financial truth.

❌ High coordination cost

❌ Transaction contention

❌ Serialization pressure

Quant Edge Exchange

Traffic and truth are separated into specialized storage layers.

✅ High-throughput ingestion

✅ Strong settlement consistency

✅ Clear separation of concerns

System Architecture

Stage 01 • Traffic

Global Bid Traffic

Regions: us-east-1 • eu-west-1 • ap-southeast-1 • sa-east-1

Purpose: Generate high-volume auction traffic.

Stage 02 • DynamoDB

Transient Traffic Layer

Stores short-lived bid events

Single Table Design

PK = SLOT#{slotId}

SK = REGION#{region}#BID#{bidId}

✓ Fast writes

✓ Regional aggregation

✓ TTL lifecycle management

Stage 03 • Evaluation

Auction Decision Engine

Ranks bids and selects the winning settlement candidate.

Stage 04 • Aurora DSQL

Authoritative Business Truth

Stores authoritative business outcomes

✓ Winning bids✓ Settlement records✓ Financial ledger✓ Conflict telemetry

Stage 05 • Analytics

Observability & Audit Layer

Tracks settlement activity, conflict metrics, and operational telemetry.

Database Responsibilities

Amazon DynamoDB

Optimized For: High-Throughput Event Ingestion.

✓ Bid ingestion

✓ Regional traffic aggregation

✓ Telemetry collection

✓ Event lifecycle management

Amazon Aurora DSQL

Purpose: Authoritative financial settlement.

✓ Winning bids

✓ Settlements

✓ Financial ledger

✓ Conflict telemetry

✓ Serializable transactions

Conflict Resolution with OCC + Full Jitter

01 // Concurrency Collision

Worker A settle(slot_123)

Worker B settle(slot_123)

↓

Serialization Conflict

02 // Retry Engine Protocol

Optimistic Concurrency Control handles cross-node clashes gracefully using an exponential backoff formula mixed with full randomized jitter:

retryDelay = random(0, base × 2^attempt)

03 // Execution History Logs

Attempt #1 ❌ Aborted

Attempt #2 ❌ Aborted

Attempt #3 ✅ Committed

Successful Settlement

PLATFORM COMPONENTS

Frontend LayerNext.js 16TypeScript / Tailwind

Runtime CoreVercel EdgeGlobal Serverless Nodes

Traffic LayerAmazon DynamoDBSingle-Table Key Partition

Truth LayerAmazon Aurora DSQLActive-Active Serializable ACID

Security AuthenticationAWS SigV4Cryptographic Request Signing

ObservabilityTelemetry DashboardCustom Live Metrics UI

Key Architectural Insight

"Not every event deserves global consistency."

Traffic Events ──► Amazon DynamoDB

Financial Truth ──► Amazon Aurora DSQL

The cost of coordination should match the value of the data.