Bazar System – Multi-tier Online Book Store

Project Description

This project implements a distributed online bookstore using a microservices architecture.

The system was first implemented in Lab 1 using three services:

• Frontend Service
• Catalog Service
• Order Service

In Lab 2, the system was extended to support:

• Replication
• Caching
• Cache consistency
• Round-robin load balancing
• Performance measurements
• Docker-based deployment

All services are implemented using Python Flask and communicate using REST APIs and JSON.

---

Lab 2 Architecture

The Lab 2 system runs five services:

• frontend
• catalog1
• catalog2
• order1
• order2

The frontend is not replicated. It receives all client requests and forwards them to backend replicas.

Client
  |
  v
Frontend
  |-- In-memory cache
  |-- Round-robin load balancer
  |
  |---- Catalog1
  |---- Catalog2
  |
  |---- Order1
  |---- Order2

---

Data Files

For Lab 2, replicated data files are stored in the data/ directory:

data/catalog1.csv
data/catalog2.csv
data/orders1.csv
data/orders2.csv

The original Lab 1 files are kept for backward compatibility:

catalog/catalog.csv
order/orders.csv

Lab 2 uses the files inside data/.

---

Services and Ports

Service	Description	Host Port
frontend	Client-facing service, cache, load balancer	5000
catalog1	First catalog replica	5001
catalog2	Second catalog replica	5003
order1	First order replica	5002
order2	Second order replica	5004

---

Main Features

Frontend Service

The frontend:

• Receives all client requests.
• Performs round-robin load balancing.
• Caches item info responses.
• Invalidates cached items when backend data changes.

Frontend endpoints:

GET  /
GET  /health
GET  /search/<topic>
GET  /info/<id>
POST /purchase/<id>
POST /invalidate/<id>
POST /cache/clear
GET  /cache/state

Catalog Service

The catalog service:

• Stores book data in CSV files.
• Supports query by topic.
• Supports query by item.
• Updates stock.
• Synchronizes updates with the other catalog replica.
• Sends cache invalidation requests before writes.

Catalog endpoints:

GET /health
GET /query/topic/<topic>
GET /query/item/<id>
PUT /update/<id>
PUT /sync/update/<id>

Order Service

The order service:

• Handles purchase requests.
• Checks stock from catalog replicas.
• Updates catalog stock.
• Stores orders in CSV files.
• Synchronizes order logs with the other order replica.

Order endpoints:

GET  /health
GET  /orders
POST /purchase/<id>
POST /sync/order

---

How to Run the Project

Make sure Docker Desktop is running.

From the project root directory, run:

docker compose up --build

To stop the system:

docker compose down

---

Health Check Commands

curl.exe http://localhost:5000/health
curl.exe http://localhost:5001/health
curl.exe http://localhost:5003/health
curl.exe http://localhost:5002/health
curl.exe http://localhost:5004/health

Expected result: all services should return "status": "ok".

---

API Examples

Search Books by Topic

curl.exe http://localhost:5000/search/distributed%20systems

Example output:

[
  {
    "id": "1",
    "title": "How to get a good grade in DOS in 40 minutes a day"
  },
  {
    "id": "2",
    "title": "RPCs for Noobs"
  },
  {
    "id": "5",
    "title": "How to finish Project 3 on time"
  }
]

---

Get Book Info: Cache MISS

First request:

curl.exe http://localhost:5000/info/2

Example output:

{
  "cache": "MISS",
  "id": "2",
  "price": 50,
  "quantity": 10,
  "served_by": "http://catalog1:5001",
  "title": "RPCs for Noobs"
}

---

Get Book Info: Cache HIT

Second request for the same item:

curl.exe http://localhost:5000/info/2

Example output:

{
  "cache": "HIT",
  "id": "2",
  "price": 50,
  "quantity": 10,
  "served_by": "http://catalog1:5001",
  "title": "RPCs for Noobs"
}

---

Purchase Book

curl.exe -X POST http://localhost:5000/purchase/2

Example output:

{
  "book_id": "2",
  "catalog_used": "http://catalog1:5001",
  "forwarded_to": "http://order1:5002",
  "message": "bought book RPCs for Noobs",
  "order_id": 1,
  "processed_by": "order1",
  "response_time_ms": 52.99,
  "title": "RPCs for Noobs"
}

After purchase, the cached item is invalidated.

---

Info After Purchase

curl.exe http://localhost:5000/info/2

Example output:

{
  "cache": "MISS",
  "id": "2",
  "price": 50,
  "quantity": 9,
  "served_by": "http://catalog1:5001",
  "title": "RPCs for Noobs"
}

This shows that the cache was invalidated and the updated quantity was fetched from the catalog.

---

Replication Checks

Catalog Replication

curl.exe http://localhost:5001/query/item/2
curl.exe http://localhost:5003/query/item/2

Both catalog replicas should return the same quantity.

Order Replication

curl.exe http://localhost:5002/orders
curl.exe http://localhost:5004/orders

Both order replicas should show the same orders.

---

Performance Experiments

The performance script is located at:

experiments/performance_test.py

Run it using:

python experiments/performance_test.py

It generates:

docs/performance_results.csv
docs/performance_summary.md
docs/response_time_plot.png
docs/cache_invalidation_plot.png

Measured results:

Operation	Mode	Average Response Time
GET /info/2	Without cache	15.72 ms
GET /info/2	With cache	4.27 ms
POST /purchase/7	Buy request	67.33 ms

Caching improvement for repeated info queries:

72.84%

---

Documentation Files

Important documentation and output files are located in docs/:

docs/design_document_lab2.md
docs/output.txt
docs/performance_summary.md
docs/performance_results.csv
docs/response_time_plot.png
docs/cache_invalidation_plot.png

---

Notes

• Each service runs in a separate Docker container.
• All communication is done through REST APIs.
• The frontend is not replicated.
• The frontend contains the in-memory cache and round-robin load balancer.
• Catalog and order services are replicated.
• Cache invalidation is performed before catalog writes.
• CSV files are used for persistence because this is a lab-scale project.