Quantum Software Architecture Design Course

Written by: Igor Korotach

Section 4

Advanced Architectural Styles & Patterns

Small recap

General approaches

Monolith Architecture

Pros

• Simplicity
• Ease of deployment
• Ease of testing
• Better single node performance

Cons

• Resource-full scaling
• Tight coupling
• Limited to a single technology

Pros & Cons of Monolith Architecture

Service-oriented Architecture (SOA)

When to assess using microservices?

• Your customer has quite a big budget
• You have non-functional requirements related to throughput 
• You can't afford vertical scaling at same time as horizontal scaling
• You have a team of diverse specialists (i.e. different tech stacks)
• You have a special case of business requirements 
• You have solid internal processes

Pros

• Improved scalability
• Flexibility of deployment
• Ease of testing

Cons

• More complexity
• Network overhead
• Harder debugging & observability

Pros & Cons of SOA

Broker pattern & Load Balancer

Broker pattern (Message Broker)

Pros

• Improved decoupling
• Potential improvements to performance
• Enhanced reliability if done correctly

Cons

• Increased complexity
• Reliance on the broker
• Worse debugging & troubleshooting

Pros & Cons of Broker Pattern

Load Balancer

Difference between Load Balancer & Broker Pattern

• Load Balancer operates on a network level, while broker pattern operates on the service level
• Main purpose of Load Balancer is enhancing the throughput, Broker Pattern is oriented towards decoupling business logic
• Load Balancers are usually stateless, while Brokers can handle context and meta information
• A good example of Load Balancer is Nginx, a good example of a Broker Pattern is Apache Kafka, Erlang actor model
• Broker Pattern is used to build more complex architecture styles such as Publish Subscribe (Pub/Sub) or Event-Driven Architecture

Pipe & Filter Pattern

Pipe & Filter Use Cases

• Text processing

• Image processing

• Streaming data

• Data analytics

• Scientific computing

• Log processing

• Content Management platforms

Since Pipe & Filter is mostly used in stream processing, below is the list of recommended tools to be used when implementing the pattern

• Apache Airflow for DAG (Directed Acyclic Graph) construction, manipulation & observability. Best used as a soft real-time tool
• Apache Flink for streamed real-time data processing
• Apache Spark for more batch-oriented highly parallelizable systems with focus on building filters
• Kafka Streaming if you already employ Kafka ecosystem

Pros

• Modularity
• Composability
• Scalability

Cons

• Increased complexity
• Network & data copying overhead
• Lack of adaptibility

Pros & Cons of Pipe & Filter

Event-Driven Architecture (EDA)

EDA Use Cases

• Chat applications

• Real-time Financial data processing

• User Activity Tracking

• Webhooks 

• Gaming

• Fraud detection

Event-Driven Architecture can be tricky to get right, I recommend decomposing it into the following pieces:

• Event Source - usually a result of parsing/scraping data periodically or a stream of events from client app
• Event Bus - usually a message broker capable of serving & storing large streams of events. Kafka/RabbitMQ/Cloud Native queues are good solutions
• Event Sink - a destination of the event, processing unit which transforms event into a business fact. Usually an asynchronous worker attached to a message queue

Pros

• Loose coupling
• Scalability
• Asynchrony

Cons

• Increased complexity
• Increased latency
• Ordering problems

Pros & Cons of EDA

Event sourcing & CQRS

Let's start with event sourcing!

State-based storing

Event-based storing

Event sourcing database

CQRS (Command-Query Reponsibility Segregation)

Event sourcing & CQRS

Event Sourcing & CQRS can be very difficult to implement, I recommend the following tools to help you with that:

• Event Store which a purposely built database for event sourcing
• Apache Kafka with permanent log retention (beware Kafka is first and foremost a message queue not a database)
• Specialized frameworks such as Axon Framework or Akka (Pekko) Persistence

Pros

• Excellent observability & bookkeeping
• Segregation of concern
• Simpler queries

Cons

• Very high complexity
• Difficult debugging
• Eventual consistency of the database
• Command Tables are very hard to modify

Pros & Cons of Event Sourcing & CQRS

Hexagonal Architecture

Pros

• Promotes testing
• Improved Modularity
• Improved Extensibility

Cons

• Increased complexity
• Requires understanding of moving parts in advance
• Can be challenging to scale horizontally 

Pros & Cons of Hexagonal Architecture

Serverless Architecture

Use Cases for Serverless Architecture

• Your customer is a small-medium sized business with limited budget
• You have received a sizable amount of Cloud credits for your provider to build PoC/MVP
• You have a predictable workload at all times
• You don't plan to have huge traffic spikes
• You have a limited development team

Pros

• Good scalability
• Low configuration costs
• Easy to setup 
• Improved reliability
• Improved security

Cons

• Limited control
• Cold-start issues
• Dependency on the cloud provider

Pros & Cons of Serverless Architecture

Next time: Scalability, Availability, Load Balancing, Caching, Security Strategies

Thanks for your attention. You've been awesome!

Questions?

• Presentation link: https://slides.com/emulebest/architecture-4