A 50-store retail network: retail, Oracle, an RDS farm and high-traffic web on the same cloud

A retail chain with 50 stores and a head office — an estate where several IT verticals run side by side. Stores side: point-of-sale terminals and the UNA accounting system. Office side: a corporate environment for 700+ concurrent users — terminal sessions, files, email, Bitrix24, printing, development. Layered on top: an in-house web platform — ten sites with combined traffic of more than 300,000 unique daily visitors, catalogue and stock exports from the accounting system, Oracle under the main load.

Before our engagement, the server estate was heterogeneous: part of it was the company's own decade-old hardware visibly running out of life, part of it leased servers from several providers — with no unified SLA, no end-to-end monitoring and no verifiable backup strategy for the critical Oracle database. Data export scripts to the web storefront ran the way they had grown over the years, and at peak hours the sites went down.

The job — pull all of this into a single, predictable engineering system with reasonable economics and transparent 24/7 support.

The full estate of the network — corporate, retail, web and dev — in our cloud, under one administration and one SLA:

• Corporate environment for 700+ concurrent users: a single domain, an RDS farm, files, printing, corporate email, Bitrix24.
• Stores and UNA: support of the accounting system and POS infrastructure across 50 stores, with network connectivity for every location.
• Oracle on Linux: a database server with proper sizing and a verifiable backup strategy.
• Web infrastructure: hosting and operation of ten websites with combined traffic of 300,000+ unique daily visitors, with a WAF at the perimeter.
• Export scripts and joint work with the client's developer team on code performance during peaks.
• Secure remote access via Zero Trust for off-site staff.
• Network: Wi-Fi on Ubiquiti UniFi, MikroTik routers, centralised monitoring.
• Hardware procurement: sourcing and delivery of 130 refurbished PCs to the client's order.

Cloud platform: migration with no cost increase

All client servers moved into our cloud from physical hardware (part of it decade-old kit on-site, part leased from several providers). The goal was not "cloud because it's fashionable" but a concrete improvement: a predictable SLA, getting out of the role of owning ageing iron, unified monitoring, and proper backups.

In our cloud the client got latest-generation servers — with NVMe disks and DDR5 memory — at essentially the same cost as the previous mix of own and leased capacity. The performance gain is most visible on Oracle and on heavy exports into the web.

Corporate environment for 700+ concurrent users

A single Active Directory domain has been deployed, with redundant domain controllers. Terminal access runs through a farm of eight RDS servers behind RDS Connection Broker: on logon, the user is distributed to the least-loaded node; if any of the eight goes down, sessions are redistributed to the rest without taking the farm out of service. Current capacity — 700+ concurrent users.

Plus: a file server with role-based access by department, a print server, corporate email, a Bitrix24 server for CRM and the corporate portal. Off-site staff connect via Zero Trust access — granted per application, without exposing the whole internal network.

Stores and UNA

The 50-store chain runs on the UNA accounting system. We operate the UNA server side, the POS infrastructure and the network connectivity from each store back to the central perimeter. At store level — Wi-Fi on Ubiquiti UniFi with a centralised controller and MikroTik routers with redundant uplinks.

Oracle on Linux

The base load of the retail and analytics circuits — Oracle Database on Linux. The server is sized for the workload; scheduled maintenance and updates run on a schedule.

The Oracle backup strategy is built on the official tool — Oracle RMAN (Recovery Manager) — and has three layers:

• Full backup (FULL) — once a week.
• Incremental backups (Level 1) — daily, on top of full.
• Archived redo logs — written continuously, enabling point-in-time recovery: the ability to restore the database to any moment within the log retention window.

Copies land in our centralised backup storage, separate from the production server; retention is set with a 30-day rolling window plus monthly snapshots kept for up to 12 months. Regularly — on the dev server — a test restore is performed: RMAN VALIDATE plus a full restore of the latest backup with archived logs applied. That gives real confidence the backup is operational — not just "taken every night on paper".

Web infrastructure and WAF

The web perimeter — ten client websites with combined traffic of 300,000+ unique daily visitors. Hosting and operation on our side. A Web Application Firewall sits in front of the sites: it filters automated attacks, common exploitation attempts and anomalous traffic; rules are tuned to each application's profile.

Joint work with the client's developer team

The websites are built by the client's own developer team; we own the platform they run on. That requires constant dialogue, not "a quarterly email".

We hand monitoring data over to the developers — slow queries, peak-hour bottlenecks, traffic and script anomalies; they optimise the code based on that. The data export scripts from the accounting system into the web storefront were rewritten jointly — and after that the sites stopped falling over at peak load.

Dev server

A dedicated dev server is provided for the developer team: isolated from production, with a separate data copy. It is used for experiments, performance testing and — among other things — for the regular Oracle backup test restores described above.

Monitoring and network security

Every key component — RDS nodes, Oracle, the file server, email, Bitrix24, web servers, routers and UniFi access points — is wired into central monitoring with on-duty alerting. Performance, availability and security metrics live on a single dashboard. That delivers early detection — not "we'll fix it once a user notices".

Hardware procurement

As part of the engagement, on the client's order, we sourced and delivered 130 refurbished PCs. That allowed a quick refresh of store and office workplaces with predictable quality and without overpaying for new hardware where it wasn't needed.

Outcome

• 700+ concurrent users work on a single infrastructure under one SLA.
• 50 stores on UNA — POS infrastructure and accounting under unified support.
• Move off decade-old hardware and leased servers into the cloud with no meaningful cost increase — onto a noticeably more performant platform (NVMe + DDR5).
• Oracle is faster than on the previous infrastructure.
• The websites stopped falling over at peak load after the joint rewrite of the export scripts.
• Oracle backups moved from "are taken" to "are taken and regularly verified by restore on the dev server".
• 10 websites at 300,000+ DAU — behind a WAF, under monitoring, with continuous performance feedback to the developers.
• 130 refurbished PCs — a fleet refresh with transparent economics.