[Architecture] System Design goals

 

System Design goals

In future, System should handle the defined non-functional requirements.

• Scalability: System can handle increased load without performance degrade
• Performance: Optimized response time, Resource utilization
• Reliability: Fault-tolerant, Data consistency, minimize failures and recover quickly
• Availability: Accessible and operational with min downtime
• Security: Protect resources/data from unauthorized access & cyber threats
• Maintainability: Easy debug, updates or feature addition
• Flexible & Extensibility: Modification & enhancements without major changes
• Cost efficiency: Balancing performance & reliability while optimizing infra costs

 

Scalability

• Vertical scaling (Scale Up): increase RAM/CPU
• Horizontal scaling (Scale Out): Add more machines to dist load
• e.g. Adding more web servers behind a LB
• Tools: K8S (Initializing Pod instead of whole VM), AWS Autoscaling, LB – Nginx, HA Proxy
• Elastic scaling (Serverless): Dynamically adjusting resources based on demand
• Don’t care about VMs/servers, Cloud provider manages on our behalf
• Tools: AWS Lambda, K8S horizontal Pod Auto-scaler

 

Best practices:

• Use MS Archi to scale individually services independently
• Implement Caching (Redis, Memcached) to reduce DB load
• Optimize DB sharding to distribute
• Use CDNs (Content Delivery Networks) e.g. CloudFare  / AWS CloudFront / Cloud CDN for faster content delivery – heavy text, images, videos etc.

 

Performance & Efficiency

• Use indexing & Caching
• Pagination, PageRank algo – pre-rank web pages
• Dist caching for Redis/Memcached to store frequent search queries
• LB & Distributed computing: Dist requests across multiple servers
• HW LB: F5 / Big IP, Citrix ADC
• SW LB: NGnix, HA Proxy

 

• Network / Layer 4 LB : operate at transport layer (TCP/UDP) – based on IP, e.g. Linux virtual servers (LVS)
• Network / Layer 7 LB : operate at application layer– NGnix, HA Proxy, Makes decision based on contents (URL, cookies, headers)

 

• Cloud based LB: offered by cloud providers, e.g. AWS Elastic LB
• DNS based LB: Uses DNS to dist traffic across multiple endpoints

 

• Optimized DS&Algo: Optimize code, index in DB, MapReduce / Hive
• HW optimization: High perf TPUs, High speed SSDs and Low latency networks