应用与设计模式

Keypoints	Architecture	Patterns
News Feed Use a fan-out / fan-in strategy for feed updates to enhance availability (1) Read-Your-Writes Consistency (1, 2) Maintain data replication across servers for consistency(1, 2, 3)	Architecture: Event-Driven Structure: Microservices	Fan-Out / Fan-In Strategy: Distribute a new post to all follower feeds (fan-out); aggregate feeds when users log in (fan-in) Event Sourcing Pattern: Store changes to the news feed as events, ensuring all writes are recorded for user consistency Replay events to rebuild user feed state in case of server failure, ensuring data consistency Sharding Pattern: Shard news feed data by user region to distribute load and reduce latency Shard by user activity level, placing more active users on high-performance shards to balance load
Share Pregenerate URLs for later use Implement caching for frequent redirects to reduce latency (1) Apply rate limiting and validation to enhance security(1)	Architecture: Request-Driven Structure: Monolithic	Cache-Aside Pattern: Cache frequently accessed short URLs to speed up redirect requests Evict expired URLs from cache to ensure users are redirected correctly Rate Limiting Pattern: Limit URL creation requests to prevent abuse and ensure fair usage Apply rate limiting on redirects to prevent DDoS attacks on popular URLs Publisher-Subscriber Pattern: Notify caching systems to evict or preload URLs when new ones are pregenerated
Location Related Implement efficient matching algorithms for low latency（1, 2） Use GPS data for location tracking Use log database to record events to facilitate data traceability	Architecture: Event-Driven Structure: Microservices	Geode Pattern: Use geodes to efficiently match riders with nearby drivers for low latency Store and index GPS data geospatially for quick location-based queries Event Sourcing Pattern: Record all ride-related events for full traceability of the system's state Use events to reconstruct a trip's state in case of disputes or audits Saga Distributed Transactions Pattern: Manage complex transactions like booking, payment, and confirmation across services Ensure consistency in distributed systems when riders cancel or change routes
IM efficient messaging protocols like XMPP (1) Ensure data consistency with eventual consistency model Encrypt messages for security (1)	Architecture: Peer-to-Peer (P2P) Structure: Monolithic	Valet Key Pattern: Give clients temporary keys to upload/download attachments securely Use keys to authenticate WebSocket/XMPP connections for sending/receiving messages
Live Optimize for real-time video delivery Use scalable cloud services, CDN Maintain stream quality with adaptive bitrate streaming using RTMPS and WebRTC (1) Thundering herd problem (1)	Architecture: Event-Driven Structure: Monolithic	Sharding Pattern: Shard video streams across multiple servers for load balancing Reduce latency by directing viewers to the nearest shard or CDN endpoint Cache-Aside Pattern: Cache popular streams and fetch from the origin server if not in cache Update cache with the most viewed streams to handle sudden traffic spikes Circuit Breaker Pattern: Prevent system overloads during peak times by cutting off non-critical operations Degrade service if a component fails to maintain overall system function
Video Use scalable cloud services like CDN Implement adaptive bitrate streaming for dynamic network conditions Use redundant systems, failover strategies Architect for horizontal scalability	Architecture: Request-Driven Structure: Monolithic	Sharding Pattern: Distribute video content across multiple servers to balance load Serve users from the nearest data shard to reduce latency Cache-Aside Pattern: Store frequently accessed videos in cache to minimize read latency Load data into the cache on demand, ensuring popular content is quickly accessible Circuit Breaker Pattern: Prevent cascading failures during peak loads by stopping operations to a failing service Enable failover mechanisms to redirect traffic if a service or server fails
File Sync Implement efficient file transfer protocols like Rsync algorithm (1) Use asynchronous and chunked file uploading for efficiency (1) Shift to microservices improved latency, resiliency, and resource efficiency(1)	Architecture: Event-Driven Structure: Monolithic	Asynchronous Request-Reply Pattern: Handle file uploads where the client initiates a request and is free to perform other tasks while waiting for the response Process large file downloads in the background CQRS Pattern: Separate read requests (retrieving file lists) from write requests (uploading files) for optimized performance Allow quick access to file metadata while ensuring consistent writes during uploads Retry Pattern & Circuit Breaker Pattern: Retry file transfers when transient network issues occur Open circuit breaker to prevent a failing file transfer service from being overwhelmed
Search Implement Inverted Index for efficient keyword matching (1) Personalize search results using user history and PageRank/Panda Optimize search algorithms for speed and accuracy	Architecture: Request-Driven Structure: Microservices	Sharding Pattern: Shard the search index by keyword ranges to distribute the load across servers Parallelize web pages across shards to enhance performance and scalability Cache-Aside Pattern: Cache popular search results and serve them quickly for repeated queries Update the cache proactively when frequent keywords are queried to reduce latency Publisher-Subscriber Pattern: Publish search queries to a topic, with subscribers updating personalized ranking asynchronously Use subscribers to trigger index updates when new content is available or changed
Ranking Design for scalability to handle complex, multi-dimensional queries Develop a caching strategy for frequently accessed filtered results Pre-compute rankings to meet specific business timeframes for updates Batch multiple update transactions into a single operation to optimize performance (1) Employ Lossy Counting and Global Query Buckets for efficient, approximate data aggregation	Architecture: Request-Driven Structure: Monolithic	Cache-Aside Pattern: Cache top tags or tracks to quickly serve frequent queries without hitting the database Sharding Pattern: Shard data by tags or music genres to distribute query load and improve scalability Allow parallel processing of multi-dimensional queries across different shards Fan-Out/Fan-In Strategy: Fan-out updates to pre-computed rankings across multiple systems, then fan-in for a global view Aggregate data from multiple shards (fan-in) for complex queries to generate final results
Trading Implement efficient data streaming services for low-latency market updates Ensure reliable trade execution with minimal processing delay Integrate timely financial news delivery and custom alerting features Facilitate secure and compliant fund transfer mechanisms	Architecture: Event-Driven Structure: Monolithic	CQRS Pattern: Separate read queries for financial news delivery from write operations for trade execution Streamline command processing to reduce trade execution latency Valet Key Pattern: Secure fund transfers by providing temporary credentials for transactional operations Minimize data access scope using valet keys for third-party financial news integration
Cooperative editing Implement operational transformation or CRDTs for low-latency, consistent collaborative editing (1) Maintain document consistency with version control and conflict resolution mechanisms (1) Design for scalability to handle simultaneous edits by thousands of users Optimize for multi-device compatibility, ensuring consistent user experience	Architecture: Event-Driven Structure: Monolithic	Event Sourcing Pattern: Record each edit as an event, enabling operational transformation for collaborative editing Replay events to resolve conflicts and maintain document version history Compensating Transaction Pattern: Reverse an edit operation when a conflict is detected to maintain consistency Apply compensating transactions to handle operational transformation conflicts
Booking Employ distributed transactions, possibly with XA protocol, for cross-service operations Prevent write skew and phantoms using serializable isolation levels in databases	Architecture: Request-Driven Structure: Monolithic	Saga Distributed Transactions Pattern: Coordinate event bookings across multiple services without XA protocol overhead Handle failures in one service by executing compensating transactions in others Event Sourcing Pattern: Store state changes as events to prevent write skew in distributed transactions Replay events to rebuild system state for serializable isolation level enforcement
Shopping Implement elastic search for efficient, scalable product discovery Employ event sourcing for reliable order tracking and management Secure user accounts with OAuth and multi-factor authentication Utilize a microservices architecture to ensure scalability of user and order systems	Architecture: Event-Driven Structure: Microservices	Event Sourcing Pattern: Record order events to track status changes Enable modifying orders by replaying and adjusting events Cache-Aside Pattern: Cache product listings and details Invalidate cache on product updates to ensure users see current information
CPU-bound Serverless architecture to reduce system complexity Utilize divide and conquer, map-reduce strategies to break down tasks Manage variable workloads with auto-scaling cloud services	Architecture: Request-Driven Structure: Monolithic	Fan-Out/Fan-In Strategy: Distribute PDF conversion tasks across multiple functions Apply video effects in parallel, then merge clips into a final video Asynchronous Request-Reply Pattern: Queue PDF conversion jobs and notify users asynchronously when done Submit video rendering jobs and provide results via callback or webhook Health Endpoint Monitoring Pattern: Monitor health of serverless functions to ensure availability for video editing Track PDF converter service health to scale or redeploy functions
IoT Utilize event sourcing to capture and store each user interaction and device event Implement efficient data compression / protocol for low-latency synchronization like MQTT Employ eventual consistency model for user data synchronization	Architecture: Event-Driven Structure: Monolithic	Event Sourcing Pattern: Record data as events for historical analysis Capture device temperature changes to recreate historical states for troubleshooting Publisher-Subscriber Pattern: Devices publish sensor data, which apps subscribe to for real-time updates Subscribe to user preferences changes to update device behavior dynamically Sharding Pattern: Shard data by user ID for scalable, distributed data storage Use geographical sharding to reduce latency for location-based IoT devices
Infrastructure Use partitioning for scalable throughput Replicate data for high availability Implement consistent hashing for consistency Encrypt data at rest for security	Architecture: Event-Driven / Request-Driven Structure: Monolithic	Health Endpoint Monitoring Pattern: Monitor the health of nodes, ensuring consistent message processing Check object storage service health to detect and resolve access issues Event Sourcing Pattern: Store every change in system configuration as an event Record log generation and access events to reconstruct system states during incidents