SAP on AWS: how about availability?
Do you know what the availability is for a single EC2 instance on AWS?
The availability is shown as 99,95%, but it is not guaranteed! That doesn’t sound good, but I will explain why this does not have to be an issue.
AWS delivers the tools and infrastructure so cheap that you can easily build a high available solution. You only have to remember that also for non-production SAP systems you need to create a basic high available solution.
For the non-production systems we can build a cheap high available solution based on an initial image and backup. After both SAP and the database is installed and the configuration is done you create an initial image called an Amazon Machine Image (AMI). This AMI can be stored in a different Availability Zone (this is a different datacenter in the same region) or even on a totally different Region. With this image you have a new instance up-and-running within 5 minutes. Besides this of course we need a restore of the data to be up-to-date. When you store your backups and archivelogs on AWS S3, it is automatically saved in multiple Availability Zones and the durability is 99,999999999%. So with this you have the cheapest solution for your non-production SAP systems.
For production systems lots of high available solution are possible, just like there are when you place your systems on-premise. The solution depends for example on the database vendor. But for production systems everyone is already aware. I just wanted to point out the risk and possible solution for the non-production systems.
Below is an example of a simple SAP architecture where the production system has a high available solution based on database replication to a second Availability Zone. For security reasons we use different subnets and different Security Groups (firewalls). The non-production systems are shown in one single Availability Zone, but we know to place the AMI’s and backups in the second Availability Zone. To be even more safe, you can not only use a second Availability Zone but even place it on another region. But this is more expensive and you have to think about laws and regulations for your data.