Kennametal backs its smart factories with SAP on Azure

Staying on top in a competitive market

Since 1938, Kennametal’s willingness to invest in new technologies has meant groundbreaking products and services for its customers. As an essential business across industries such as aerospace, energy, transportation, and general engineering, companies turn to Kennametal to help them manufacture with precision and efficiency. Headquartered in Pennsylvania, Kennametal employs nearly 10,000 employees in more than 60 countries, and the business generated approximately $2.4 billion in revenue in the past year.

Like many companies, Kennametal has relied on SAP ERP software for years. Before moving to Azure, critical SAP ERP systems were hosted on IBM Power System servers in the company’s local datacenter in Latrobe, Pennsylvania. These systems kept track of the manufacturing schedules, plans, and output. Less-critical applications ran in regional datacenters.

Over the years, the IT landscape had grown increasingly complex and expensive to manage, and it required a proliferation of vendors for support. The global IT team couldn’t scale the infrastructure or provision new services fast enough to meet the changing demands of the business.

"We were investing heavily in smart factory components, but our processes were still mostly manual," says Dan Berlin, Kennametal Global IT Director, Business Integration, User Experience and eCommerce. "We lacked intelligence within the factory floor, and consuming data and using it to drive decisions was a very time-consuming process."

Digital transformation takes shape

As part of its Simplification/Modernization overall strategy, Kennametal looked to the cloud to meet its long-term strategy to optimize IT infrastructure, operations, and maintenance. “We were looking for the right platform to create the factory of the future,” notes Chandra Madugula, Senior Manager of Infrastructure Delivery and Operational Excellence (IDOE) at Kennametal. “We needed a new architecture and a move from on-premises to the cloud.”

The company wanted to improve business continuity, as well. Historically, third-party recovery of Kennametal’s business systems took up to 72 hours. The team knew it could pare that number down. Even before the move to Azure, it optimized the disaster recovery (DR) exercises to less than 24 hours. But with a cloud-based architecture, Kennametal hoped to set real-time objectives for disaster recovery, backup, and restoration times.

Kennametal’s experience with other Microsoft products put Azure in the running. The company was an early adopter of Office 365 and other cloud services, including Dynamics CRM and Power BI. These tools had given employees the productivity edge they needed, whether they worked on the factory floor, in research, or in the executive suite.

“We looked at the capabilities we needed for the factory of the future and the integration that was required with our current technology stack,” explains Berlin. “The partnership with Microsoft and the great integration across products that Microsoft has—these were key factors in our decision to go with Azure.”

Another factor was the Microsoft relationship with SAP. At the time, Microsoft was in the process of getting the Azure infrastructure certified by SAP to run its workloads. Today, Azure holds certifications for multiple SAP products, and both Microsoft and SAP use Azure to run mission-critical SAP applications. Recently, the two companies deepened their relationship when SAP selected Azure as its preferred partner to help customers reduce complexity, minimize costs, innovate their businesses, and reach new levels of digital transformation.

While waiting for the SAP certification, Kennametal proceeded to migrate its non-SAP workloads to Azure in a straightforward lift and shift. The next step was to migrate applications that required refactoring or updates, followed by applications with licensing issues. These early projects gave the team the confidence it needed to tackle the SAP ecosystem.

Kennametal worked closely with an Azure cloud architect to document the capacities and server loads of its existing stack to determine the new architecture on Azure.

“Once we dug into the details and could begin to forecast the costs, we had a lot more confidence,” Berlin says. “We could see how the cloud enabled us to pull levers you don’t think about on-premises, like shutting down dev systems you don’t need or scaling up only at peak times.”

The architecture of a smart factory

In Kennametal’s main on-premises datacenter, the SAP application tier ran on IBM Power Systems server frames. For the move to Azure, the team needed to replatform legacy assets and move them to Azure Virtual Machines running either Windows or SUSE Linux.

To do this, a fundamental architectural difference had to be resolved. The IBM systems store bytes in computer memory using an approach known as big-endian, and systems based on an x86 architecture use a different approach, known as little-endian. For four months, the team explored various migration approaches. The solution was the SAP Software Provisioning Manager (SWPM) tool. The team used it to migrate the SAP ERP workloads to Azure—an effort that also required moving approximately 75 interfaces.

The move consolidated Kennametal’s three main datacenters located in the United States, Europe, and Asia. Cost reduction was a goal, but network latency is also a key consideration when migrating SAP workloads. The SAP application tier requires a clear line of sight to the data tier to support frequent communications. In reviewing the most cost-effective connectivity solutions, the migration team considered multiple options, weighing the benefits of regional proximity and the cost of third-party colocation sites.

The new architecture hosts the SAP ERP systems in the Azure East US datacenter. The other regional services moved to various global datacenters operated by a colocation service provider. An Azure ExpressRoute circuit connects the Azure infrastructure to the Kennametal global network. “We had an old network,” Berlin notes. “In moving to the cloud, we had to ask, is our infrastructure built for that? We revamped our entire network to be cloud-ready.”

Kennametal evolved its network architectures to take advantage of Azure capabilities and to provide connectivity to external networks used by internal users, customers, and partners. Azure Virtual Network service guarantees a natural boundary of isolation for Kennametal resources and is set up in a hub-spoke topology. The hub is used for shared services, and the spoke virtual networks (VNets) are used for applications and other services.

Virtual network peering supports intercommunication between different VNets within the same Azure region, across regions, or even between VNets in different subscriptions. Inside a VNet and between VNets, Kennametal controls traffic flows using security rules defined in network security groups, firewall policies, and network virtual appliances.

To estimate infrastructure requirements for each SAP system, the team looked at historical trends and used SAP Application Performance Standard (SAPS), a hardware-independent unit of measurement that describes the performance of a system configuration in the SAP environment. The application tier uses memory optimized D-series and E-series virtual machines (VMs).

Azure Load Balancer distributes traffic to VMs in the application tier subnet. For high availability, VMs are deployed using availability sets that distribute the load across multiple isolated hardware clusters. “Azure VMs simplified our process of building and decommissioning systems as we went,” says Madugula.

The data tier uses the M-series family of VMs, the largest memory-optimized VMs to date. The M-series is designed for heavy, in-memory workloads, such as SAP HANA, and offer up to 4 TB of RAM on a single VM.

For business continuity and disaster recovery, Kennametal uses Azure Site Recovery, which provides native disaster recovery as a service (DRaaS). Site Recovery replicates workloads running in the Azure East US region to a secondary location hosted in the Azure West US region. Before Azure, the team had improved its recovery time to 24 hours, but Kennametal targeted a bolder objective for DR, backup, and restoration—three hours. The Azure cloud-based architecture meets that goal.

What it takes to migrate fast

To minimize disruption to the business, the team agreed to a two-day cutover window. To meet this ambitious goal, the team created a detailed, 60-day plan and coordinated closely with all its software vendors to test application communications. All the teams took part in multiple full-dress rehearsals, during which critical system integrations were tested. Design reviews were held with Microsoft and SAP to identify potential risks. Daily standups, dashboards, and scorecards held teams accountable throughout.

The careful planning paid off. “We carried out multiple iterations and tests with our partners and IT teams until we could do the actual ERP migration in less than 25 hours,” says Madugula. During the cutover window, Kennametal received added support from Azure Rapid Response teams in various regions, who provided round-the-clock help.

“We wanted more bandwidth and high network availability during the migration. We had to build multiple WAN channels with high bandwidth that could be torn down easily,” explains Rajul Kuttiadi, the network architect who headed the migration network team. “Additionally, Azure Traffic Manager helped us to minimize time for DNS propagation to ensure a seamless migration.”

Kennametal used Azure Site Recovery extensively as a migration tool and afterward as the key to its disaster recovery strategy. The rest of the migration was automated as much as possible through Azure Resource Manager scripts. Another valuable tool was Movere, a discovery solution that helps in planning a cloud migration by scanning and capturing everything in an IT environment.

In the data tier, the Oracle database was migrated to Microsoft SQL Server running on VMs, while a few non-SAP databases were migrated to Azure SQL Database, a fully managed service. To move the associated 3 TB of data within the two-day cutover window, the team used several approaches. For example, to avoid bottlenecks, the team used table splitting during the export process and parallel rather than sequential load processes during the import process.

Security from the ground up

The new architecture takes advantage of a variety of Azure services. “Security was a big piece of the puzzle,” says Berlin. “Microsoft provides a much higher level of security for our data and infrastructure within Azure than we could ever provide in our internal datacenter.”

For example, the IT team can secure its data more easily than it could on-premises. Azure Backup takes snapshots of the data according to a backup schedule. Azure Blob storage also takes snapshots that capture the blob state at that point in time. In addition, SQL Server Agent automates and schedules backups. “The automated SQL Server backups help us recover our systems faster,” notes Madugula. “They’re also easier to configure and manage than our old system.”

Other Azure features helped the team build a secure framework from the ground up. The team used Azure resource groups to isolate production workloads from non-production resources. The VNets also support isolation. The network topology is designed for distributed applications and teams and to support subnet delegation, which helps with secure network management. Even the Azure storage accounts provide security—they use role-based access control (RBAC) to grant access to Azure resources as needed.

An architecture of change

The SAP architecture on Azure boosted the company’s analytics efforts and has inspired more Azure projects. “We have our SAP system in Azure. We have our analytics warehouse in Azure,” explains Brad Zorn, Kennametal Director of Digital Manufacturing and Enterprise Analytics. “Now we are collecting machine data in Azure from our factories, and we can very quickly write cross-database queries that allow us to combine all this information together.”

For example, welding machines in one factory emit data that is sent to a server, scrubbed, and then sent to Azure, where it is stored in SAP databases. Power BI dashboards aggregate this data and combine it with information from other factory machines. Ultimately, the company hopes to feed the machine data into the SAP ERP system and use it to optimize the routes and the scripts that automate plant operations.

In a proof-of-concept experiment, the company began pulling the material specifications from its ERP system and comparing it to data captured from the factory floor, such as the room’s temperature and humidity. Using Azure Machine Learning, automated solutions can analyze this information and begin to predict changes in a factory’s operations and to improve efficiency.

Summary

Simply by consolidating the regional datacenters, Kennametal met a long-term goal to optimize its IT infrastructure and lower the cost of maintenance. Now the IT team can scale up and down based on changing business requirements or try out new technologies and features with easy configuration tweaks. New applications can be deployed in days or weeks—instead of months—and global teams can develop new solutions using a consistent set of tools and services and a common portal.

Most importantly, the critical SAP systems run better overall on the scalable infrastructure that Azure provides. Azure has also improved the company’s cybersecurity stance by helping to ensure the integrity and confidentiality of the data.

But Azure also opened the door to a wealth of new opportunities. Kennametal admits that it didn’t entirely know what it had in Azure until after the migration was complete, when it began experimenting with machine learning, artificial intelligence, and other Azure options.

Now the company is letting the data tell it where to go next. “We started asking, what can we do with these tools to quickly get a glimpse of all this data coming out of our factories? The speed with which you’re able to do experiments and then scale—that’s incredibly powerful in the Azure platform,” says Zorn.