Back to Blog

What I Learned at AWS About Managing SLOs

All we could do was apply automation to fix the issues when they occurred, providing customers much better availability.

You should run your service as a utility company.

Here’s what I mean:

Back at AWS, we used to describe what we did as utility computing – where you could just plug in and get compute, storage, database, and 150 other things.

It’s how you get power and gas from your local utility company – metered, on-demand, scales with you, etc.

Now, you want a lot of things from your utility company. But mostly, what you want is that the lights stay on at your place. You don’t care if they have five-nines power availability across the state.

That’s why it's important to take a customer-centric approach, particularly when you're managing data services.

It’s because you might be able to move from one piece of compute to another, but you can't move from one database to another net new database that doesn’t have your data.

So at AWS, we used to ticket per individual database instance.

By the time I left, there were like 5 million of those things, many of which were created by someone else, where we couldn't change the software.

So all we could do was apply automation to fix the issues when they occurred, providing customers much better availability.

That was part of why our margins grew 60% year over year while I was there.

Now, at Shoreline, we develop automations that fix issues forever within a few hours (maybe twice the time it takes you to fix it once manually) instead of months.

Thus, we help our customers improve availability for each of their customers.