Skip to Content
Product Information
Author's profile photo Limor Wainstein

Running SAP HANA on Azure Spot Instances

What is SAP HANA?

SAP HANA is an in-memory database and application development platform used to process large volumes of information in real-time.

SAP HANA lets data analysts query sizable volumes of information in real-time. HANA’s in-memory computing means no time is wasted loading or writing data to disk, allowing analysts to perform queries much faster. 

HANA’s columnar-based information store is atomicity, consistency, isolation, and durability (ACID)-compliant. HANA supports industry standards including multidimensional expressions (MDX) and structured query language (SQL).   

SAP HANA also features a programming element that lets an organization’s IT department develop and run customized applications on top of HANA and a suite of predictive, text, and spatial analytics libraries over multiple information sources. HANA may run in parallel to a source SAP ERP application. Given this, analysts may access real-time transactional and operational data for real-time analytics processing without waiting for a weekly or daily report to run.  

Why Run SAP HANA in Azure?

You can run SAP HANA as an Azure cloud database, hosted on Azure infrastructure. This has several advantages:

Lower CAPEX

Migration to SAP HANA in the cloud implies simplicity of scaling via the employment of their remote servers. Scaling down is also simple. According to research from Forrester, the average organization can save $7.2 million in hardware costs by transferring SAP applications to Azure (rather than upgrading their data centers).

Automated Disaster and Backup Recovery Systems

Azure Site Recovery that smoothly integrates with SAP can safeguard all your critical assets by replicating the disks. Furthermore, you may develop system replication between the on-site and cloud instances to safeguard your database. You can also build single- or multi-region disaster recovery systems based on your needs. These systems can help you maintain the desired RTO and RPO Values.  

Improved Data Sharing and Workflows 

SAP in the cloud encourages efficient collaboration without geographical restrictions and permits real-time file-sharing from any device and any place. Furthermore, the Microsoft IoT hub now smoothly integrates with SAP Leonardo, so you can launch more sophisticated IIoT deployments and gain real-time insights from multiple physical assets. 

Integrated Offerings, Supporting Large Scale Deployments

Azure provides several very large instances to support huge SAP HANA deployments. Specifically, they made Azure Mv2 Virtual Machines (VMs) available with as much as 6 TB of memory for SAP HANA. Furthermore, customers may select custom SKU configurations for as many as 480 Intel CPU cores and as much as 24 TB of memory. 

What Are Azure Spot VMs?

Azure spot instances are an Azure option that lets you utilize the platform’s unused capacity. Users get Azure Spot VMs at an exceptionally discounted rate – this feature can help Azure users reduce cloud costs by up to 90%

Azure Spot VMs do not guarantee high availability. If Azure evicts your Spot VM, Microsoft will not alert you unless you specify that you wish to receive in-VM notifications through Azure Scheduled Events. If Microsoft requires capacity or the spot prices go beyond your maximum price, the eviction notification time is only 30 seconds. Therefore, a typical use case for a Spot VM is an application or service where interruptions are permissible. 

If, for example, you have a workload that utilizes multiple GPUs or CPU-based computers to complete complex mathematical algorithms, the Azure Spot VM pricing model offers a balance between price and performance. Since these tasks generally include extended periods in which calculations are waiting to finish, an interruption will not affect the user experience. It could take a bit longer, but because users do not access the service directly, this does not have a production impact.

However, with SAP HANA, as a mission critical workload, any interruption could be critical. A possible solution is to deploy HANA in scale-out mode, with multiple workers connected to a master. The master node should be deployed on a regular Azure VM, with high availability. The workers can be deployed on spot instances, and if they are terminated, they can be replicated from the master.

Deploying SAP HANA on Azure Spot Instances in Scale-Out Mode

The following is a brief about deploying an SAP HANA database via Azure Spot Instances:

Connect Into SAP Virtual Machines on Azure

Before you launch an SAP HANA database via a VM, you must organize a method of connecting to the VM from your on-site data center. 

You may connect via a public endpoint via Jump VM. Alternatively, you can connect directly through the Internet to the VM deploying SAP HANA. However, this may create security issues.

A safer way for production situations is to connect through a VPN or a specific Azure ExpressRoute connection.

Select Azure VM Types

After choosing an Azure VM type that Microsoft permits for use via HANA, there are various methods of deploying the VM. You can employ the Azure CLI, Azure portal, or Azure PowerShell commands. Alternatively, you can use SAP Cloud Platform to launch and run managed SAP HANA instances in an Azure VM. 

You can decide to employ an Azure spot instance when running the VM, as depicted in the screenshot that follows. Choose Yes under Azure Spot instance, and then select from these spot VM choices: 

  • Eviction types—state whether you are willing to pay a greater price than your first bidding price for the spot instances (Capacity only) or if you choose to terminate the VM if the price goes above your bid (Price or capacity).
  • Eviction policy—states what should occur when Azure terminates the spots instance—if you wish to stop the instance, with the option to restart, or delete it entirely. 

Image source: Azure

Choose View pricing history to contrast spot VM pricing with different Azure regions.

Handling Interruptions

An Azure spot instance may be interrupted with 30 seconds notice. To be prepared for this, run HANA in scale-out mode, as shown in the diagram below. The architecture is founded on Azure NetApp Files (ANF) and utilizes high-performance ANF file shares to distribute information between several HANA instances. 

You may run some of the worker nodes via spot instances, and then if Microsoft interrupts them, the other workers can adopt the load. Ensure that the master node is deployed on a regular instance so that it won’t be interrupted. 

Image source: Azure

Conclusion

In this article, I explained the benefits of running SAP HANA in the Microsoft Azure cloud, and explained how running HANA on spot VMs can result in dramatically lower operating costs. 

I provided a sample implementation that lets you:

  • Run multiple SAP HANA instances on Azure spot VMs
  • Leverage NetApp Azure Files as a high-speed file system
  • Enable connectivity between the on-premise data center and HANA in Azure
  • Deal with interruptions to spot VMs by running HANA in a scale-out model, and ensuring the master node is running on a regular Azure VM (not a spot VM)

I hope this will be of help as you explore more optimal ways to run HANA databases in the Azure cloud.

Assigned Tags

      1 Comment
      You must be Logged on to comment or reply to a post.
      Author's profile photo Jan Tretina
      Jan Tretina

      Hi, interesting approach to use Azure spot instances, however we need to mention that it is not supported and certified scenario, one should always use only standard Azure VMs with SKUs listed in the SAP Note 1928533.