Relationships (association, aggregation, composition) in LLD - Scalability & System Analysis

Scalability Analysis - Relationships (association, aggregation, composition)

Growth Table: Relationships in System Design

Users / Objects	100	10,000	1,000,000	100,000,000
Number of Relationship Links	~1,000	~100,000	~10,000,000	~1,000,000,000
Storage for Relationship Metadata	MBs	GBs	TBs	Petabytes
Query Complexity	Simple joins	Moderate joins, indexing needed	Complex joins, caching required	Distributed joins, sharding mandatory
Latency Impact	Negligible	Noticeable without optimization	High without caching and sharding	Critical, requires advanced optimization
Data Integrity Checks	Simple constraints	Automated constraints, triggers	Distributed consistency challenges	Eventual consistency, complex coordination

First Bottleneck

The first bottleneck is the database handling relationship data. As relationships grow, join queries become expensive. Especially for composition where lifecycle dependencies require strict integrity, the database load increases. Without proper indexing and caching, query latency spikes.

Scaling Solutions

• Indexing: Add indexes on foreign keys to speed up joins.
• Caching: Cache frequent relationship queries to reduce DB hits.
• Read Replicas: Use replicas to distribute read load.
• Sharding: Partition data by user or object to reduce join scope.
• Denormalization: Store some relationship data redundantly to avoid complex joins.
• Eventual Consistency: For aggregation, relax strict consistency to improve performance.
• Lifecycle Management: For composition, use transactional or distributed transactions carefully.

Back-of-Envelope Cost Analysis

Assuming 1 million users with 10 relationships each:

• Relationships: 10 million links
• Storage: ~100 bytes per link -> ~1 GB total
• Queries: 10,000 QPS (e.g., 10,000 active users each querying once per second)
• Bandwidth: 1 KB per query -> ~10 MB/s network traffic
• Database: Needs to handle 10,000 QPS, likely requiring sharding and replicas

Interview Tip

Start by explaining the types of relationships and their impact on data integrity and lifecycle. Then discuss how relationship volume affects database load. Finally, propose scaling strategies like indexing, caching, and sharding, explaining why each fits the problem.

Self Check

Your database handles 1000 QPS for relationship queries. Traffic grows 10x. What do you do first?

Answer: Add read replicas and implement caching to reduce direct database load before considering sharding or denormalization.

Key Result

Relationship data first stresses the database with complex joins and integrity checks; indexing, caching, and sharding are key to scaling.

Practice

1. Which relationship describes two objects connected without ownership, where both can exist independently?

easy

A. Inheritance

B. Association

C. Composition

D. Aggregation

Solution

1. Step 1: Understand relationship types

   Association means objects are linked but do not own each other and can exist separately.

2. Step 2: Compare with other relationships

   Aggregation and composition imply ownership; inheritance is a different concept.

3. Final Answer:

   Association -> Option B

4. Quick Check:

   Association = connected without ownership [OK]

Hint: No ownership means association, not aggregation or composition [OK]

Common Mistakes:

• Confusing aggregation with association
• Thinking composition allows independent existence
• Mixing inheritance with relationships

2. Which of the following correctly shows a composition relationship in UML notation?

easy

A. A solid diamond at the owner end connected to the part

B. A hollow diamond at the owner end connected to the part

C. A plain line with no diamonds between objects

D. A dashed arrow pointing from part to owner

Solution

1. Step 1: Recall UML notation for composition

   Composition uses a solid diamond at the owner end to show strong ownership.

2. Step 2: Differentiate from aggregation and association

   Aggregation uses hollow diamond; association uses plain line; dashed arrow is for dependency.

3. Final Answer:

   A solid diamond at the owner end connected to the part -> Option A

4. Quick Check:

   Composition = solid diamond [OK]

Hint: Solid diamond means composition, hollow means aggregation [OK]

Common Mistakes:

• Mixing hollow and solid diamonds
• Confusing association line with composition
• Using dashed arrows for composition

3. Consider this code snippet representing relationships:

class Engine:
    pass

class Car:
    def __init__(self):
        self.engine = Engine()

What type of relationship does Car have with Engine?

medium

A. Association

B. Aggregation

C. Inheritance

D. Composition

Solution

1. Step 1: Analyze ownership in code

   Car creates and owns Engine instance inside its constructor, implying strong ownership.

2. Step 2: Match with relationship definitions

   Since Engine cannot exist meaningfully without Car here, this is composition.

3. Final Answer:

   Composition -> Option D

4. Quick Check:

   Car owns Engine fully = Composition [OK]

Hint: Creating owned object inside means composition [OK]

Common Mistakes:

• Thinking it's aggregation because of ownership
• Confusing association with composition
• Assuming inheritance from code

4. Identify the error in this aggregation example:

class Wheel:
    pass

class Car:
    def __init__(self, wheels):
        self.wheels = wheels

car = Car([Wheel(), Wheel()])

What is wrong with this aggregation?

medium

A. Wheels cannot exist independently in aggregation

B. Car should not hold wheels as a list

C. No error; this is correct aggregation

D. Wheels should be created inside Car for composition

Solution

1. Step 1: Understand aggregation ownership

   Aggregation means Car owns wheels but wheels can exist independently.

2. Step 2: Check code for ownership and independence

   Wheels are created outside Car and passed in, so they can exist alone, matching aggregation.

3. Final Answer:

   No error; this is correct aggregation -> Option C

4. Quick Check:

   Aggregation allows independent parts [OK]

Hint: Aggregation parts created outside owner, passed in [OK]

Common Mistakes:

• Thinking wheels must be created inside Car
• Confusing aggregation with composition
• Assuming list usage is wrong

5. You are designing a library system where a Library contains many Books. Books can exist outside the library and be moved between libraries. Which relationship best fits this scenario and why?

hard

A. Aggregation, because library owns books but books can exist alone

B. Composition, because books depend fully on the library

C. Association, because library and books are loosely connected

D. Inheritance, because books inherit properties from library

Solution

1. Step 1: Analyze ownership and existence

   Books can exist independently and move between libraries, so they are not fully owned.

2. Step 2: Match relationship type

   Aggregation fits because library owns books but books can exist alone and be shared.

3. Final Answer:

   Aggregation, because library owns books but books can exist alone -> Option A

4. Quick Check:

   Aggregation = ownership with independent parts [OK]

Hint: If parts can exist alone, use aggregation [OK]

Common Mistakes:

• Choosing composition thinking books depend fully
• Selecting association ignoring ownership
• Confusing inheritance with ownership