chat

High Level Requirements

Functional

• User Authentication

• One-to-One Chat

• Group Chat

• Online Presence

• Read/Delivered Status

• Typing Indicator

• Push Notification

• Media Upload

• Message History

• Search Conversation

Non Functional

• High Availability (99.99%)

• Horizontal Scalability

• Low Latency (<100ms)

• Eventual Consistency

• Fault Tolerance

• Secure Communication

• Multi Region

• Millions of concurrent connections

Capacity Estimation

فرض کنیم

100 Million Users

10 Million Daily Active Users

1 Million Concurrent Connections

Average Message Size = 1 KB

100 Messages/User/Day

Traffic

10M ×100

=1 Billion Messages/Day

≈11,600 Message/sec Average

Peak
≈100,000 Message/sec

بنابراین طراحی باید حداقل

100K Write/sec

را پاسخ دهد.

High Level Architecture

                    +----------------+
                    | Mobile/Web App |
                    +-------+--------+
                            |
                     HTTPS / WebSocket
                            |
                  +---------v---------+
                  | Global LoadBalancer|
                  +---------+---------+
                            |
        +-------------------+------------------+
        |                                      |
+-------v------+                     +---------v-------+
| API Gateway  |                     | WebSocket Gateway|
+-------+------+                     +---------+--------+
        |                                      |
        |                                      |
        |                         Persistent Connections
        |                                      |
+-------v--------------------------------------v------+
|               Chat Service Cluster                  |
+--------+---------+----------+-----------+-----------+
         |         |          |           |
         |         |          |           |
         |         |          |           |
+--------v+   +----v----+ +---v----+ +----v-----+
|Presence |   |Message  | |Group   | |Media     |
|Service  |   |Service  | |Service | |Service   |
+----+----+   +----+----+ +----+---+ +----+-----+
     |             |           |          |
     |             |           |          |
     |      +------v-----------v----------+
     |      | Kafka / Event Bus           |
     |      +-------------+---------------+
     |                    |
+----v-----+        +-----v------+
|Redis     |        |Notification|
|Cluster   |        |Service     |
+----+-----+        +-----+------+
     |                    |
     |               APNS / FCM
     |
+----v----------------------------+
| Cassandra / Scylla / DynamoDB   |
+---------------------------------+

Media
  |
S3 / Object Storage

C4 Context Diagram

                +----------------------+
                |      User            |
                +----------+-----------+
                           |
                           |
                    Uses Chat System
                           |
                +----------v-----------+
                |   Chat Platform      |
                +----------+-----------+
                           |
      +--------------------+--------------------+
      |                    |                    |
      |                    |                    |
+-----v----+        +------v-----+      +-------v------+
|Push       |        |Object      |      |Identity      |
|Provider   |        |Storage     |      |Provider      |
|(FCM/APNS) |        |(S3)         |      |OAuth/JWT)    |
+-----------+        +------------+      +--------------+

C4 Container Diagram

                     +------------------+
                     | API Gateway      |
                     +--------+---------+
                              |
      +-----------+-----------+-----------+
      |           |           |           |
      |           |           |           |
+-----v--+  +-----v---+ +-----v---+ +-----v-----+
|AuthSvc |  |ChatSvc  | |GroupSvc | |PresenceSvc|
+-----+--+  +-----+---+ +-----+---+ +-----+-----+
      |           |           |           |
      +-----------+-----------+-----------+
                      |
                Kafka Cluster
                      |
       +--------------+---------------+
       |                              |
+------v------+               +-------v------+
|Notification |               |Analytics     |
+-------------+               +--------------+

              |
      +-------v--------+
      | Cassandra      |
      +----------------+

Sequence Diagram (Send Message)

User A

 |
 | Send Message
 |
 v

Gateway

 |
 | Authenticate
 |
 v

Chat Service

 |
 | Persist Message
 |
 v

Database

 |
 | OK
 |
 v

Kafka

 |
 +----------------------------+
 |                            |
 v                            v

Notification             Presence Service

 |                            |
 |                            |
 v                            |
Recipient Online?             |
 |                            |
 |Yes                         |
 |                            |
 v                            |
WebSocket Push                |
                              |
                              |
                         Recipient Offline
                              |
                              v
                         Push Notification

Entity Diagram

+-------------+
| User        |
+-------------+
| id          |
| username    |
| status      |
| lastSeen    |
+------+------+
       |
       |1
       |
       |N
+------v------+
|Conversation |
+-------------+
| id          |
| type        |
| createdAt   |
+------+------+
       |
       |1
       |
       |N
+------v------+
| Message      |
+--------------+
| id           |
| senderId     |
| content      |
| type         |
| sentAt       |
| status       |
+------+-------+
       |
       |
       |N
       |
       |1
+------v------+
| Attachment  |
+-------------+
| id          |
| url         |
| size        |
| mimeType    |
+-------------+

Conversation

1

|

N

ConversationMember

N

|

1

User

Database Choice

DataStorageMessagesCassandra / ScyllaUser ProfilePostgreSQLPresenceRedisCacheRedis ClusterMediaS3SearchElasticsearch

Kafka Topics

message-created

message-delivered

message-read

user-online

user-offline

typing-event

notification-event

media-uploaded

Cache Strategy

Redis

Online Users

Conversation Cache

Unread Counter

Recent Messages

Session Token

Rate Limiter

Sharding Strategy

Conversation Based

Shard = hash(conversationId)

مزایا

• 
  تمام پیام‌های یک گفتگو روی یک شارد
  

• 
  Query ساده
  

• 
  ترتیب پیام‌ها حفظ می‌شود
  

Message Flow

Client

↓

Gateway

↓

Chat Service

↓

Kafka

↓

Database

↓

Redis Update

↓

Notification

↓

Receiver

ADR (Architecture Decision Records)

ADR-001

Decision

استفاده از WebSocket برای ارتباط Real-Time.

Reason

• 
  Full Duplex
  

• 
  Latency پایین
  

• 
  کاهش هزینه نسبت به Polling
  

ADR-002

Decision

ذخیره پیام‌ها در Cassandra/ScyllaDB.

Reason

• 
  Write Throughput بسیار بالا
  

• 
  Horizontal Scaling
  

• 
  Partitioning داخلی
  

• 
  مناسب Time-Series Data
  

ADR-003

Decision

استفاده از Kafka برای Event Streaming.

Reason

• 
  Loose Coupling
  

• 
  Retry
  

• 
  Replay
  

• 
  Back Pressure
  

• 
  Event Driven Architecture
  

ADR-004

Decision

استفاده از Redis برای Presence.

Reason

• 
  عملیات In-Memory
  

• 
  TTL
  

• 
  سرعت بسیار بالا
  

ADR-005

Decision

Media در Object Storage.

Reason

• 
  کاهش حجم دیتابیس
  

• 
  CDN Integration
  

• 
  ارزان‌تر
  

• 
  Scalability بالا
  

ADR-006

Decision

Horizontal Scaling برای Chat Service.

Reason

• 
  Stateless Service
  

• 
  Auto Scaling
  

• 
  Load Distribution
  

Failure Scenarios

FailureSolutionChat Service DownLoad Balancer + Auto ScalingKafka DownReplication + ISRRedis DownRedis Cluster + ReplicaDatabase Node DownMulti ReplicaGateway FailureMultiple Gateway InstancesRegion FailureMulti-Region DeploymentMessage DuplicationIdempotency KeyMessage OrderingPartition by ConversationId

Trade-offs

ChoiceProsConsWebSocketReal-time، تأخیر کممدیریت Connection دشوارCassandraنوشتن بسیار سریع، مقیاس‌پذیرJoin محدود، Consistency نهاییKafkaجداسازی سرویس‌ها، تحمل بارپیچیدگی عملیاتیRedisسرعت بالاداده‌ها عمدتاً در حافظهMicroservicesمقیاس‌پذیری و استقلالپیچیدگی شبکه و مانیتورینگEvent-Drivenانعطاف‌پذیری و توسعه‌پذیریDebug و Trace دشوارتر

این طراحی در سطح مصاحبه‌های Senior Java Developer / Software Architect جامع است و تقریباً تمام نمودارهای کلیدی مورد انتظار (High-Level Architecture، C4 Context، C4 Container، Sequence Diagram، Entity Diagram، ADR، سناریوهای خطا و Trade-offها) را پوشش می‌دهد.