The cloud is changing IT best practices in such a fundamental way, that transforming On Premise or Hosting Landscapes to the cloud requires re-thinking to get the best results. To be clear, very often On Premise and Hosting solution architectures could be migrated one-to-one including the approaches to HA. This will however not lead to significant improvements over prior state. Customers come to the cloud to:

• increase de-facto Business Uptime
• increase de-facto Security and Data Protection
• increase Business Innovation
• lower cost

Maximum Uptime

Not everyone has 2+ clusters of data centers in a GEO region and in each of them blades with 2 TB RAM on 4-socket Intel® Xeon® E7 8880 v3 --- IDLE, just waiting to be used by some LOB.Well that is exactly what AWS provides. AWESOME.

And with Ocean9 a SAP HANA Cluster providing a 99.99% uptime utilizing 3 Data Centers in one GEO region can be provisioned in 35 minutes using a single button click.

The architecture above has one downside, though. A triple site architecture leads to triple infrastructure cost.

The single-site Cluster

The term "single-site cluster" is a contradiction for most experts. How can we call a deployment with no (obvious) redundancy in the architecture Highly Available and use even the term "Cluster" ?

Well in fact this architecture is used by many enterprises around the world to run any SAP solution (not just SAP HANA) in production on AWS. The reasons why this makes sense in a large scale cloud and it does not make sense elsewhere are:

• In a large scale cloud you can have with a high probability any amount of infrastructure at any location any time. With the right solution architecture and operations you can increase that probability to close to 100%.
• Many of the AWS platform services are inherently designed for maximum uptime. Leveraging those services as foundation for SAP HA, enables you to sleep at night.
• With AWS you pay only for what you use, not for what you could use.

So what happens when one Availability Zone (= cluster of datacenters) fails?

Well you recover and failover into another data center in the same AWS GEO region. For SAP HANA Ocean9 provides a recovery time of less than 2 hours and a maximum risk of data loss of 15 minutes.(RPO: 15 minutes / RTO: 2 hours)

And today we are making this available with the simplicity of a button click. We are calling it "Anytime Failover".

Anytime Failover

When you have the need to failover to another site, and there is no secondary site to failover to, we first provision a secondary site, recover the latest backup and failover into it, immediately after.

And as the name indicates, the button Anytime Failover will also work if you have a Dual or Tripe Site HANA Cluster with SAP HANA System Replication. In that scenario Ocean9 will follow the replication trail and re-configure the networking accordingly.

Summary

The advantage of a Single Site Cluster is Infrastructure cost optimization. It is literally AWS cost neutral. It does not cost you a single infrastructure dollar extra compared to a Standalone System.

Compared to a Standalone System, a Single Site Cluster helps you also with

• Hardware Refreshes (no data loss,no downtime)
• SAP HANA Patches / Upgrades (no data loss,no downtime)
• Scale Up (no data loss,no downtime)

It is therefore just the better system.

In addition you can always add a site with the click on "Add Site" and thereby increasing your Uptime tier (1-->2 Sites or 2-->3 Sites) in no more than 30 minutes. 

As always, we hope you liked it.

Continue Reading:

• Taking High Performance to SAP Mission Critical Systems
• SAP HANA on AWS Configurator
• SAP HANA with 2 TB RAM on AWS