Home / server / Questions / 1030152
Accepted
zcahfg2
Asked: 2020-08-15 08:24:01 +0800 CST

AWS API gateway average downtime per year per region?

  • 772

Our team is trying to see if we should use the managed AWS API gateway over IDC Kong for API Gateway.

So I am looking for some concrete statistics for AWS managed service (In particular, AWS API Gateway).

But the total downtime per year region statistics is surprisingly difficult to find from the AWS official website.

I looked at the SLA agreement (https://aws.amazon.com/api-gateway/sla/) that seems to guarantee up to 95% availability - that is totally out of the question. 18 days per year we might as well just go out of the business. And they only reimburse the credit used - not the total losses incurred due to the downtime.

There is also service health dashboard that shows current health and status history. But I need something way better than that - pie chart for total downtime in minutes during 2019, 2018, 2017 in each regions.

enter image description here

I found this stats "Modernize your Microsoft Applications on AWS" but again its not specific enough - it combines all services and regions.

Why are these stats so difficult to find? I am sure this is one of the more important metrics for choosing the service?

Many thanks in advance

amazon-web-services statistics

1 Answers

  1. Best Answer

    That SLA document says the SLA is 99.95%, not your stated 95%.

    AWS will use commercially reasonable efforts to make API Gateway available with a Monthly Uptime Percentage of at least 99.95% for each AWS region.

    The 95% is only mentioned in the table that shows how much credit you get back if they breach the SLA.

    Anyway, things break and services go down. But not all at the same time.

    Your API GW may break (on its own or due to your actions), but perhaps not all API GWs at once. Availability Zone may go down but perhaps not all AZs at the same time. Yes sometimes even some Region goes down but other Regions are fine.

    If your service is super critical and you can’t afford an outage make it highly-available through geo-redundancy. That means deploy the same API in different regions, perhaps one region in US, one in EU and one in Asia-Pacific.

    AWS provides a lot of tools to do that, from unified deployment through CloudFormation templates, through DynamoDB Global tables, to latency based routing with Route53. And more.

    Single instance / API GW uptime / etc doesn’t matter if your deployment is HA.

    Hope that helps :)

    • 2