🌐 Wrapping Up: Scaling Systems, Load Balancers, and High-Level Design

Stateful Systems and Context Dependency

Even though all machines in a stateful cluster run identical code, they hold different sets of data.
For instance:
- Machine A may store conversation data for User 101.
- Machine B may store similar data for User 107.
Subsequent requests from User 107 must return to Machine B for context-based responses, creating session stickiness.

Clusters and Code Uniformity

A cluster comprises machines running the same code but catering to specific data needs.
Example:
- Search Service Cluster: Machines run code for search operations.
- Product Service Cluster: Machines run product-related logic.

This structure ensures efficient scaling and resource utilization across distributed systems.

From High-Level Design (HLD) to Implementation

HLD focuses on architecting scalable systems—understanding distributed computing, gateways, and load balancers.
Practical implementations (e.g., projects) allow applying this knowledge:
- You may use cloud services for prototyping while leveraging HLD principles for system integrity.
- Not every concept will apply to a single project; learning spans years of experience and diverse applications.

Where Does the Firewall Reside?

The firewall is part of the gateway layer, serving as a first-line defense.
It filters incoming traffic based on predefined rules, ensuring that only legitimate requests reach backend services.

Active-Passive Configuration

A load-handling strategy:
- The active machine processes requests.
- The passive machine waits on standby and takes over if the active machine fails.
Focuses on failover management rather than data duplication.

Replication

Refers to duplicating data across machines for redundancy or data integrity.

IP Sunsetting Process

Sunsetting involves gradually deprecating an old IP (e.g., IP1) in favor of a new IP (e.g., IP2):
- DNS maintains records for both IPs during the transition.
- Over time, cached requests to IP1 reduce, allowing its eventual removal.

Static IP and ISP Dependencies

Static IPs are tied to the Internet Service Provider (ISP).
Portability of static IPs across ISPs is under exploration but not widely available yet.

Reverse Proxy vs. Gateway

A reverse proxy manages requests by forwarding them to appropriate backend servers, acting as a shield.
The gateway serves as a specialized reverse proxy, combining:
- Request filtering (security).
- Traffic forwarding.

Load Balancer as a Reverse Proxy

Load balancers often act as reverse proxies to distribute requests among backend machines.

Pre-Built Configurations

Most load balancers and gateways come with pre-written configurations.
Developers may need to write custom logic or plug-ins for specific routing rules or balancing algorithms.

Stateless vs. Stateful Load Balancing:
- Stateless: Requests can go to any machine.
- Stateful: Requests depend on specific machines for contextual data.
Scaling Strategies:
- Horizontal Scaling: Adding machines to handle increasing traffic.
- Vertical Scaling: Increasing the capacity of individual machines.
High-Level Design:
- Focuses on scalability, reliability, and distributed architectures.
- Helps make informed decisions when implementing real-world systems.
Real-World Implementations:
- Utilize cloud services for rapid deployment.
- Apply HLD principles for robust and future-proof designs.

Next, we’ll explore:

Consistent Hashing: Solving the challenges of dynamic scaling in stateful load balancing.
CDN Design: How distributed systems manage global content delivery efficiently.
Microservices: Breaking down monoliths into scalable, independent units.

These learnings form the cornerstone of mastering system design at scale.

Stay tuned as we continue to build on these foundational concepts!

Last updated 3 months ago