Skip to content
Blotout Blotout Blotout
Dashboard

Architecture

Blotout Cloud runs as a single-tenant deployment inside a private AWS VPC. The diagrams below show how infrastructure components connect at deploy time and how event data flows from collection through the data lake to analytics and activation.

Deployment

Section titled “Deployment”

Cloud Deployment Architecture — Terraform Cloud, Cloud UI, and single-tenant AWS VPC

Terraform Cloud provisions one AWS VPC per tenant. Users and analysts reach the stack through the Cloud UI (Cloudflare-hosted) over an HTTPS API. Infrastructure is defined in Git and applied by Terraform Cloud with a remote state file (tfstate) and remote execution — no local apply against production.

Inside the VPC, services are grouped into four bands:

App & orchestration — dashboard-backend (Spring Boot), dbt for transformations, Airflow (MWAA) for workflow orchestration, Amazon Bedrock for AI insights, and EdgeTag for CAPI and tag delivery.

Data & cache — PostgreSQL (RDS) as the primary datastore (dual datasource: blotout + airbyte), Redis (ElastiCache) for distributed cache, Caffeine as local L1 cache, Amazon S3 as the data lake, Athena for serverless SQL on S3, AWS Glue as the data catalog, and Firehose (Kinesis) for streaming ingest.

Segment activation (highlighted band) — DynamoDB holds segment state, SQS FIFO queues activation jobs, and EKS runs the activation workers that push audiences outbound.

Security & ops — IAM roles and policies, CloudWatch alarms and logs, Auto Scaling for EKS pods, and private zones for network isolation.

Semantic Layer

Section titled “Semantic Layer”

Data pipeline — Sources through Empower, with enterprise capabilities

Data moves left to right through seven stages, from collection to activation. Shared enterprise services under the pipeline keep the stack orchestrated, secure, and observable.

Sources — Advertising (Google Ads, Meta, TikTok, LinkedIn), commerce and payments (Shopify, Stripe), databases/files (Postgres/Files), and client-side Web/App SDK events.

Ingest — Airbyte, REST API, EdgeTag events, and Firehose pull or stream data into the platform.

Raw store — Landing zones on Clickstream S3 and Offline S3, plus source metadata, organized through the AWS Glue Catalog.

Transform — dbt models, Athena SQL, and ID stitching clean and enrich raw data into analytics-ready tables.

Data models — Curated models for Clickstream, Google Ads, Meta Ads, TikTok Ads, and integrated Stripe–Shopify commerce data.

Reporting — ID Graph, sessionization, and a Unified View that joins identities and events into a single analytical surface.

Empower — Consumption layer: Analytics Hub (Cloud dashboards), AI Insights (Amazon Bedrock), Segments (audience builder), and Activate (reverse-ETL back to operational tools).

Enterprise capabilities (foundation) — Orchestration with Airflow (MWAA); Caffeine + Redis caching; 1st-party infra, IAM, and private zones; CloudWatch monitoring and Auto Scaling.

Section titled “Related documentation”