Enterprise infrastructure

High-Volume Marketplace Software

Technologies and infrastructure requirements for marketplace sites that must support millions of products and thousands of concurrent visitors.

Load Balancer
RabbitMQ
Redis
Elasticsearch

What does “high-volume marketplace” mean?

When millions of products can be listed and thousands of customers may visit the site at once—like Trendyol, N11, or Hepsiburada—the software must be built to handle that load.

Why do you need advanced technologies?

Sellers
Server
Visitors

In a marketplace model, many sellers continuously upload and update products while large numbers of visitors use the site at the same time. That requires major improvements in both application architecture and server configuration.

The improvements below must be scaled continuously with extra servers and supporting software as product and visitor volumes grow.
1

Load Balancer

Visitors
Load Balancer
Server 1
Sunucu 2
Load balancing

A system that routes customers visiting the marketplace to the most available server.

Monitors workload across servers
Directs each visitor to an available server
Requires at least one load balancer and two servers

Cluster architecture

The two backend servers mirror each other and run the same site. Products uploaded or orders taken on one server are processed in sync on the other.

2

RabbitMQ (message queue)

Job queue
Processed
Queuing

On high-volume sites, heavy operations from sellers stress the servers and create simultaneous load spikes.

When that load exceeds what the server can handle, the site slows down, may lock up, and can temporarily go offline.

Operations that should be queued:

1
Seller product uploads
2
Product export via Excel
3
Order email notifications
4
Shipping API integrations
5
SMS notifications
3

Using Redis

User request
Redis (RAM)
Very fast
or
MySQL (Disk)
Slow
Fast response
In-memory cache

Redis is an enhancement that keeps a copy of data normally stored in the database on disk also in the server’s RAM for faster access.

MySQL (database)

Disk-based, slower query responses

VS
Redis (RAM)

Memory-based, ultra-fast responses

When values such as price or stock change, the cached data in RAM must be cleared so the latest data can be written again.

4

Elasticsearch

Elasticsearch
Instant results
Search engine

Technology that serves search results from Elasticsearch instead of SQL when users search by keyword.

Fast search across very large catalogs
Better performance when using filters
Data hosted on a dedicated server
Support for near real-time data updates

Filter performance

Combined with Redis, serving filter results from Elasticsearch helps avoid slowdowns.

Conclusion and recommendations

All of the features described above should be strengthened in line with growth—typically by adding more servers.

Example starter setup

1 Load Balancer
2 Application server
1 Elasticsearch server
1 Redis server
Correct networking between servers and reliable data synchronization across them are critical.

Free consultation

Talk to our specialists for detailed information about marketplace software and benefit from free consultation on your project.

Demo