Removing the single point of failure in a young SaaS product

Everything runs on one server, and a big new customer is asking about uptime. Two app servers and a managed database fix the real risk; a full rewrite would not.

2 min readMade-up stories, not client work

Cloud Architecture: story 7 of 15

A made-up story

Picture a small software company whose whole product runs on one server. A large customer wants to sign, but first asks what happens if that server fails.

What's going wrong

Honestly, everything stops. An adviser suggests rebuilding the product as many small services on a complex platform.

What they have today

  • One server running the app, database and background jobs
  • Manual updates
  • A team of a few developers

What we can't ignore

  • The big customer needs an answer soon
  • The team can't stop building features
  • The budget is modest

The choices

  • Not for them

    Rebuild as many small services on a complex platform

    Months of work and a platform the team would need to learn and run.

  • Worth a look

    Package the app in containers on a managed service

    A reasonable step, but more change than the problem needs right now.

  • Our pick

    Two app servers and a managed database

    If one server fails, the other carries on, and the database has a standby.

  • Not for them

    Keep one server and improve backups

    Better than nothing, but the product still stops if the server does.

What we'd do

Keep the app as it is. Move the database to a managed service with automatic failover. Run two app servers behind a load balancer, give background jobs their own small server, and automate updates. Write the setup down as code and test a failover.

Why

  • Removes the real single point of failure
  • The team keeps shipping features
  • A tested failover gives the customer a straight answer

Where the money really goes

No prices here: this shows where the costs sit, not what they'd be for any real business.

Two servers and a managed database cost more each month than one server. The rewrite would cost far more in time, and losing the big customer would cost more still.

Hosting
A second server, a load balancer and a managed database.
Engineering time
Days to weeks, not months.
The deal
An honest uptime answer helps close it.

Running, growing, and staying safe

If something breaks

If one app server fails, the other keeps serving. The database fails over automatically, and that's been tested.

When it grows

Add a third app server when traffic grows.

Keeping it safe

Automated updates and servers set up from code make it harder for mistakes to slip in.

What the business gets

  • No single box that takes the product down
  • A clear answer for big customers
  • No pause on new features

These are the kinds of results the approach aims for, not measured results.

When we'd look again

If the product grows into several teams working on separate parts, splitting out specific pieces would be worth a look then.

If you liked this, read

About these stories. These stories are made up to show how we think. They are not client projects, describe no real business, and contain no measured results.

Next: Connecting systems for a multi-location clinic group

Appointments, billing and reminders live in separate systems. A few small, well-watched connections do the job; a real-time data platform would be far more than needed.

Read the next story