Stop Paying for Scale You Don't Have: How We Cut Orbi’s Infrastructure Costs by 95%

When building a startup, runway is everything. Every dollar spent on bloated infrastructure before you have actual paying customers is a dollar you can't spend on building the product or acquiring users.

During our long-term partnership with Orbi (orbi.cz), we audited their pre-launch backend infrastructure and found a massive financial leak. Before a single real user had even touched the platform, their server costs had ballooned to $1,400 a month.

They were paying for massive enterprise scale they simply did not need yet.

Our engineering team at Yolk Studio stepped in, audited the architecture, and completely overhauled their infrastructure. In just 15 days, we reduced their monthly server bill from $1,400 to roughly $70—a 95% reduction—freeing up over $15,000 a year that Orbi immediately redirected into product development and marketing.

Here is exactly how we stripped away the enterprise bloat and built a lean, pragmatic foundation for their launch.

The Problem: Pre-Launch Enterprise Bloat

When we looked under the hood of Orbi’s backend, we found an architecture built for a massive corporation, not an agile startup.

The application was running on .NET and relying on an expensive, self-hosted SQL Server instance. Furthermore, every single server was running Windows, which meant Orbi was bleeding cash just on operating system licensing fees.

To make matters worse, they had spun up separate, expensive cloud servers for every minor sub-service (a CMS, an additional data warehouse, their ELK logging stack, etc.), complete with high-availability power links and enterprise firewalls. Because they were running a testing environment alongside production, every single one of these bloated costs was doubled.

Finally, the architecture was fundamentally mismatched to their analytics needs. They needed to run data analysis on a complete copy of the application database, but many modern data analysis services don't play nicely with SQL Server. This forced them to run an additional PostgreSQL database just for analytics.

They were paying the "enterprise tax" before they had enterprise revenue.

The Investigation: Making Pragmatic Trade-offs

We needed to right-size the infrastructure for the startup phase. This meant making calculated, pragmatic decisions about where to spend money and where to be scrappy.

1. The Database Engine: Ditching SQL Server The biggest financial drain was SQL Server. A cloud-managed instance was costing them hundreds of dollars a month, and a self-hosted perpetual license would have cost around $18,000.

• The Choice: We decided to completely migrate the database engine from SQL Server to a managed, open-source PostgreSQL instance. Not only is PostgreSQL drastically cheaper, but it also natively integrated with their required data analysis tools, eliminating the need to run two separate databases.

2. The Logging Stack: A Scrappy ELK Workaround Orbi needed the ELK stack (Elasticsearch, Logstash, Kibana) for system logging and monitoring. A managed cloud ELK instance is expensive but comes with built-in authorization. A self-hosted ELK stack is free but has no native security, meaning anyone on the internet could potentially access their logs.

• The Choice: We opted for the free, self-hosted route but implemented a smart, startup-appropriate workaround. We secured the Kibana dashboard behind an Nginx reverse proxy and configured an IP access list. This ensured the logs were only accessible from the Yolk Studio office IP address, keeping the system secure without paying enterprise licensing fees.

3. Application Hosting: Consolidating Services Rather than paying premium prices for specialized App Platforms with UI dashboards, we moved the application to standard Virtual Private Servers (VPS/Droplets).

• The Choice: We consolidated multiple smaller services onto a single, robust droplet, using Nginx as a reverse proxy to route traffic. We replaced the expensive App Platform UI by setting up automated scaling with Docker and seamless CI/CD pipelines via GitHub Actions.

The Solution: A Lean, Mean Migration

With the decisions made, our engineering team executed the migration seamlessly:

1. The Infrastructure Move: We migrated the entire setup to a more reasonably priced cloud provider (DigitalOcean). We configured the new VPS instances, set up external droplet firewalls, and reconfigured the DNS routing.

2. The Database Overhaul: We spun up the new managed PostgreSQL database and implemented pgBouncer for efficient connection pooling.

3. The Data Migration: We updated the backend ORM driver to use EF Core with Npgsql, fixed all the database migrations, and safely transferred the existing data using pgLoader.

Most importantly, we executed this entire transition with zero downtime using blue-green deployment strategies.

The Business Impact: Right-Sizing for Growth

We don't believe in doing things cheaply just for the sake of being cheap. We believe in building the right architecture for the right phase of the business.

By stripping away the unnecessary enterprise bloat, we reduced Orbi's infrastructure costs by a staggering 20x. The entire engineering effort took roughly 15 man-days, meaning the project paid for itself in less than three months.

Orbi stopped burning cash on idle servers and Microsoft licenses, and instead redirected that capital exactly where a pre-launch startup should: into building a better product and funding the marketing campaigns that actually drive growth.

If your server bill is giving you anxiety before your product has even found its footing, it might be time for an infrastructure audit.