Online presence system in HLD - Architecture Diagram

System Overview - Online presence system

An online presence system shows if users are currently active or offline in real time. It must handle many users updating their status frequently and provide quick updates to friends or contacts.

Key requirements include low latency updates, scalability to millions of users, and fault tolerance to avoid losing presence data.

Architecture Diagram

User Clients
   |
   v
Load Balancer
   |
API Gateway
   |
Presence Service <-> Cache (Redis)
   |
Database (NoSQL)

Components

User Clients

client

Users' devices sending presence updates and requesting friends' presence

Load Balancer

load_balancer

Distributes incoming user requests evenly to API Gateway instances

API Gateway

api_gateway

Handles authentication, routing requests to Presence Service

Presence Service

service

Processes presence updates and queries, interacts with cache and database

Cache (Redis)

cache

Stores recent presence states for fast read/write access

Database (NoSQL)

database

Stores durable presence history and user metadata

Request Flow - 8 Hops

User Clients → Load Balancer

Load Balancer → API Gateway

API Gateway → Presence Service

Presence Service → Cache (Redis)

Presence Service → Database (NoSQL)

Presence Service → API Gateway

API Gateway → Load Balancer

Load Balancer → User Clients

Failure Scenario

Component Fails:Cache (Redis)

Impact:Presence reads become slower as cache misses increase; writes still succeed but with higher latency

Mitigation:Presence Service falls back to database reads; cache is restored from DB asynchronously; use cache replication for high availability

Architecture Quiz - 3 Questions

Which component handles user authentication before processing presence data?

A Load Balancer

B Presence Service

C API Gateway

D Cache (Redis)

Design Principle

This design uses a cache to provide fast access to frequently changing presence data, reducing load on the database and improving user experience with low latency. The layered approach with load balancer and API gateway ensures scalability and security.

Practice

1. What is the primary purpose of an online presence system in a chat application?

easy

A. To track if users are currently online or offline

B. To store chat message history permanently

C. To encrypt messages between users

D. To manage user account passwords

Solution

1. Step 1: Understand the role of presence system

   An online presence system tracks user activity to know if they are online or offline.

2. Step 2: Differentiate from other chat features

   Message storage, encryption, and password management are separate features not handled by presence systems.

3. Final Answer:

   To track if users are currently online or offline -> Option A

4. Quick Check:

   Presence system = user online status [OK]

Hint: Presence means tracking user online/offline status [OK]

Common Mistakes:

• Confusing presence with message storage
• Thinking presence handles security features
• Mixing presence with account management

2. Which event is NOT typically used in an online presence system to track user status?

easy

A. message_send

B. connect

C. disconnect

D. heartbeat

Solution

1. Step 1: Identify presence tracking events

   Presence systems use connect, heartbeat, and disconnect to track user activity.

2. Step 2: Recognize unrelated events

   message_send relates to sending chat messages, not presence tracking.

3. Final Answer:

   message_send -> Option A

4. Quick Check:

   Presence events exclude message sending [OK]

Hint: Presence tracks connection, not message sending [OK]

Common Mistakes:

• Assuming message events track presence
• Confusing heartbeat with message_send
• Ignoring disconnect event importance

3. Given this simplified presence update code snippet, what will be the user's status after 10 seconds?

user_last_seen = 0
current_time = 10
heartbeat_interval = 5
if current_time - user_last_seen <= heartbeat_interval:
    status = 'online'
else:
    status = 'offline'

medium

A. Error due to comparison

B. "online"

C. "offline"

D. "unknown"

Solution

1. Step 1: Calculate time difference

   current_time - user_last_seen = 10 - 0 = 10 seconds.

2. Step 2: Compare with heartbeat interval

   10 <= 5 is false, so status is set to 'offline'.

3. Final Answer:

   "offline" -> Option C

4. Quick Check:

   Time diff > heartbeat means offline [OK]

Hint: Compare last seen time difference with heartbeat [OK]

Common Mistakes:

• Mixing up less than and greater than
• Forgetting to subtract last seen time
• Assuming status is always online

4. Identify the bug in this presence update logic:

def update_status(last_seen, current_time, heartbeat=10):
    if current_time - last_seen > heartbeat:
        return 'online'
    else:
        return 'offline'

medium

A. Function should return a boolean, not string

B. The comparison sign should be reversed

C. Heartbeat value should be negative

D. No bug, logic is correct

Solution

1. Step 1: Analyze the condition meaning

   If time difference is greater than heartbeat, user should be offline, not online.

2. Step 2: Correct the comparison

   The condition should return 'offline' when difference > heartbeat, so the comparison sign must be reversed.

3. Final Answer:

   The comparison sign should be reversed -> Option B

4. Quick Check:

   Time diff > heartbeat means offline [OK]

Hint: Online if time diff ≤ heartbeat, else offline [OK]

Common Mistakes:

• Returning online when user is inactive
• Misunderstanding heartbeat meaning
• Ignoring return type correctness

5. You need to design an online presence system for a messaging app with millions of users. Which approach best ensures scalability and real-time accuracy?

hard

A. Store presence data in user profile tables updated once per day

B. Store user status in a centralized database and query it on every client request

C. Send presence updates only when users log in or log out, ignoring heartbeats

D. Use distributed in-memory cache with TTL and heartbeat updates from clients

Solution

1. Step 1: Evaluate centralized database approach

   Querying a centralized DB for every request causes high latency and bottlenecks at scale.

2. Step 2: Consider distributed cache with TTL and heartbeats

   This approach keeps presence data fresh, reduces DB load, and supports real-time updates efficiently.

3. Step 3: Analyze other options

   Ignoring heartbeats or updating once per day leads to stale presence info, unsuitable for real-time apps.

4. Final Answer:

   Use distributed in-memory cache with TTL and heartbeat updates from clients -> Option D

5. Quick Check:

   Distributed cache + heartbeat = scalable real-time presence [OK]

Hint: Use cache with TTL and heartbeats for scalable presence [OK]

Common Mistakes:

• Relying on centralized DB for real-time presence
• Ignoring heartbeat updates causing stale data
• Updating presence too infrequently