Foundations: Keeping It Alive — Platform & Ops

What you'll learn

• Observability (Monitoring and Alerting)

• The Three Pillars

  | Pillar | What | Tool Examples | |--------|------|--------------| | Metrics | Numerical time-series data (CPU, latency, error rate) | Prometheus, Datadog, CloudWatch | | Logs | Structured event records | ELK Stack, Splunk | | Traces | R…

• Monitoring Architecture (Pull-Based)

  Key metrics to monitor: - RED method (request-focused): Rate, Errors, Duration - USE method (resource-focused): Utilization, Saturation, Errors

• Alerting Pyramid

• SLO, SLI, and SLA

  | Concept | Definition | Example | |---------|-----------|---------| | SLI (Service Level Indicator) | The actual measured metric | 99.3% of requests completed in Control Plane Components: - API Server -- The frontend to the cluster. All…

• Pods

  A Pod is the smallest deployable unit in Kubernetes. It represents one or more containers that share the same network namespace (same IP address) and storage volumes. Key characteristics: - Containers within a Pod communicate via localho…

• ReplicaSets

  A ReplicaSet ensures a specified number of Pod replicas are running at all times. If a Pod fails, the ReplicaSet controller creates a replacement. You rarely create ReplicaSets directly -- they are managed by Deployments.

• Deployments

  A Deployment is a higher-level abstraction that manages ReplicaSets and provides declarative updates to Pods. What Deployments add over raw ReplicaSets: - Rolling updates (gradually replace old Pods with new ones) - Rollback capability (…

• Services

  Pods are ephemeral and get new IP addresses when recreated. A Service provides a stable network endpoint (IP + DNS name) that routes traffic to a set of Pods selected by labels.

• Kubernetes Networking Model

  Kubernetes enforces a flat networking model with these fundamental rules: 1. Pod-to-Pod: Every Pod can communicate with every other Pod without NAT (Network Address Translation) 2. Node-to-Pod: Every Node can communicate with every Pod w…

• Self-Hosted vs Managed Kubernetes

  | Aspect | Self-Hosted | Managed (KaaS) | |--------|-------------|-----------------| | Setup | You provision VMs, install K8s, configure networking | Provider handles all infrastructure | | Maintenance | You patch OS, upgrade K8s version…

• Web Application Tier Architecture

  Application architecture can be classified by how its components are physically separated into distinct deployment units (tiers). This is different from layers (logical code separation) -- tiers represent actual physical/network separation.

• Single-Tier (1-Tier)

  The entire application -- UI, business logic, and database -- runs on a single machine. The user installs and runs everything locally. Examples: Desktop apps (MS Office, GIMP), mobile games with local saves, embedded systems Pros: No net…

• Two-Tier (Client-Server)

  The application is split into a client (typically UI + some logic) and a server (database or backend logic). Examples: Traditional desktop apps connecting to a shared database, thick-client banking applications Pros: Centralized data man…

• Three-Tier

  The most common web architecture. Separates concerns into three distinct tiers: - Presentation Tier -- Browser or mobile app; handles UI rendering - Application Tier -- Server-side logic; processes business rules, handles requests - Data…

• N-Tier

  Real-world systems extend beyond three tiers by adding specialized layers: Common additional tiers include: - Cache tier (Redis/Memcached) between app and database - Message queue tier (Kafka/RabbitMQ) for async processing - Search tier…

• Layers vs Tiers

  | Concept | Layer | Tier | |---------|-------|------| | Separation | Logical (within code) | Physical (separate machines/processes) | | Communication | Function/method calls | Network calls (HTTP, TCP, gRPC) | | Example | MVC pattern in…

• Saga Pattern — Distributed Transactions Without 2PC

  In microservice architectures, a business transaction often spans multiple services, each with its own database. Since 2PC is slow, blocking, and doesn't scale, the Saga pattern is the preferred alternative. Core idea: Break a distribute…

• Notification Platform Design

  A notification platform is a set of microservices that sends communication messages to users across multiple channels (SMS, Email, Push Notification, Webhooks, In-App).

• Goals

  - Fire and forget — async processing with automatic retry for intermittent failures - Ease of integration — internal services onboard through a standard API with API key authentication - Multi-channel — a single notification request can…

• Architecture

• Key Design Decisions

  1. Priority Queues: Different clients get different priority levels. An OTP message from authentication service gets higher priority than a promotional campaign notification. Use separate queues or priority headers. 2. Vendor Fallback St…

• Data Model

• Non-Functional Requirements

  - Availability: 99.99% uptime (notifications are often time-sensitive) - Latency: 4T - Score = (Satisfied + Tolerating/2) / Total samples Distributed Tracing: Tracks a single request as it flows through multiple microservices. Each servi…

• Cloud Infrastructure Fundamentals

• Regions and Availability Zones

  Cloud providers organize infrastructure in a hierarchy designed for both performance and fault tolerance: | Concept | Definition | Design Impact | |---------|-----------|---------------| | Region | A cluster of data centers in a geograph…

• Choosing Multi-AZ vs Multi-Region

  | Strategy | Protects Against | Trade-off | |----------|-----------------|-----------| | Single AZ | Nothing (dev/test only) | Cheapest, simplest | | Multi-AZ (same region) | Hardware failure, power outage, natural disaster in one AZ | S…

• Document Database Replication (Replica Sets)

  In document databases, replication uses a replica set pattern — a group of database processes hosting the same data: | Node Type | Role | Data? | |-----------|------|-------| | Primary | Receives all write operations; records them in the…

• Document Database Sharding

  When a single replica set reaches hardware limits (RAM, CPU, disk), sharding distributes data across multiple replica sets: | Component | Purpose | |-----------|---------| | Router (mongos) | Routes queries to the correct shard(s); clien…

• Database Cursors (Server-Side vs Client-Side)

  A cursor is a mechanism for iterating over a large result set without loading all rows into memory at once. When a query returns millions of rows, the choice of cursor type determines where those rows are buffered and how they flow to yo…