docs update after infrastructural change, removal of ALB from the system design

docs/architecture/00-baseline/README.md

@@ -8,7 +8,7 @@
 
 A URL shortener accepts a long URL and returns a short, human-shareable alias that redirects to the original. The system's primary job is redirection — it is not a metadata store, analytics platform, or user management system. Redirects are perceived as part of normal page navigation; any latency is user-visible. The system must therefore be optimised heavily for the read (redirect) path, while keeping the write (create) path simple and correct.
 
-The deployment target is `us-east-1` (single region). The API and short URL redirects are served at `go.buffden.com` (ALB → EC2 → Spring Boot); the Angular SPA is hosted at `tinyurl.buffden.com` (S3 + CloudFront). v1 establishes a production-minimal baseline. v2 evolves it to handle higher scale and reliability requirements without changing the deployment model.
+The deployment target is `us-east-1` (single region). The API and short URL redirects are served at `go.buffden.com` (EC2 → Spring Boot); the Angular SPA is hosted at `tinyurl.buffden.com` (S3 + CloudFront). v1 establishes a production-minimal baseline. v2 evolves it to handle higher scale and reliability requirements without changing the deployment model.
 
 ---
 
@@ -102,7 +102,7 @@ These decisions are locked before v1 implementation begins. Changing them requir
 
 Goal: ship the smallest correct system.
 
-- Components: DNS → Load balancer → Nginx → App servers → PostgreSQL
+- Components: DNS → Nginx → App servers → PostgreSQL
 - No cache required in v1; DB handles up to 5K QPS peak
 - Rate limiting at Nginx level (coarse per-IP throttle on `POST /api/urls`)
 - No soft delete required (optional `is_deleted` field, deferred to v2)

docs/architecture/00-baseline/adr/ADR-002-scaling-approach.md

@@ -19,7 +19,7 @@ The system is read-heavy (99:1 read/write ratio in v1, 95:5 in v2). The redirect
 
 ## Decision
 
-**v1**: Scale application servers **horizontally** behind a load balancer. Keep the database as a **single PostgreSQL primary** (vertical scaling + indexing). No cache required.
+**v1**: Scale application servers **horizontally** behind Nginx. Keep the database as a **single PostgreSQL primary** (vertical scaling + indexing). No cache required.
 
 **v2**: Add **Redis cache-aside** on the redirect path (> 90% cache hit target) to absorb the 20K QPS peak without overloading the DB. Enable app autoscaling on CPU + P95 latency. Add a DB read replica only if cache miss pressure on the primary becomes measurable.
 
@@ -28,7 +28,7 @@ The system is read-heavy (99:1 read/write ratio in v1, 95:5 in v2). The redirect
 ## Consequences
 
 - Application instances must be **fully stateless** — no in-process session or shared memory.
-- Load balancer and health checks are required infrastructure from v1.
+- Nginx and health checks are required infrastructure from v1.
 - The trigger for adding read replicas is a **measured metric**, not an upfront decision.
 - v2 introduces Redis as a new operational failure domain; the redirect path must degrade gracefully when Redis is unavailable (fall back to DB, apply stricter rate limiting).
 

docs/architecture/00-baseline/poc-requirements-baseline.md

@@ -80,7 +80,7 @@ Version 1:
 
 - Client (browser).
 - DNS Resolution.
-- Load balancer (L4 or L7).
+- Nginx (TLS termination + reverse proxy).
 - Stateless application servers.
 - Storage system (short -> original mapping) SQL.
 - Optional in-memory cache for hot keys.
@@ -153,7 +153,7 @@ Reason:
 
 - Read-heavy workload (80/20).
 - Stateless redirect logic.
-- App instances can scale independently behind load balancer.
+- App instances can scale independently behind Nginx.
 - DB initially vertically scaled, later read-replicated if needed.
 
 Avoid premature sharding at this stage.
@@ -166,8 +166,7 @@ Avoid premature sharding at this stage.
 | ------------- | ------------------ | ---------------------- | ---------------------------- |
 | DB            | Slow query         | Redirect latency spike | Proper indexing              |
 | DB            | Down               | All redirects fail     | Replication + failover       |
-| App           | Crash              | Partial 5xx errors     | Multiple instances behind LB |
-| Load Balancer | Misroute           | Partial outage         | Health checks                |
+| App           | Crash              | Partial 5xx errors     | Multiple instances, health checks |
 | DNS           | Resolution failure | Full outage            | DNS caching + low TTL        |
 | Cache         | Miss storm         | DB overload            | Rate limiting + TTL          |