How International Tractors Bulldozed Technical Debt with Pragmatic Virtualisation While Saving Money

International Tractors Limited, founded in 1969, is an Indian manufacturer of tractors and agricultural machinery. With two brands, Sonalik and Solis, the company has a strong presence in over 120 countries. It operates an integrated tractor manufacturing facility in Hoshiarpur, Punjab, and also has 51 skill development training centres across the country to provide farmers training opportunities on proper usage of tractors and farm implements.

The company faced two significant challenges. First, the accumulation of technical debt over time resulted in slow systems and prolonged turnaround times for reports and sales related processes, leading to user dissatisfaction about IT systems in general. Second, the management of its IT infrastructure was inadequate, as the company frequently responded to rising workloads by investing in additional hardware without a thorough assessment of actual needs. This reactive approach not only resulted in overspending but also raised potential compliance issues due to mismanaged licences.

Dhananjaya P, VP-IT, Digital Strategy & Excellence at International Tractors decided to overcome the two challenges by optimising the performance of key applications and  transitioning to virtual machines (VMs).

Streamlining Applications

“We were dealing with typical sales and service applications, all of which were custom-built in-house several years ago and were encountering performance issues. For instance, generating a dealer sales report or model mix report for a 30-day period was taking as long as 25 minutes. It was time to set the house in order,” says Dhananjaya.

To address these challenges, Dhananjaya and his team undertook a comprehensive effort over two months to enhance the applications. They assigned two capable team members to take responsibility for improving the system and ensure accountability and focus.

The development team discovered that many of the performance issues were rooted in outdated code and inefficient database queries that consumed significant processor time.

“By refactoring code and optimising database queries, we were able to address these underlying problems. The team leveraged an existing tool, ManageEngine, to analyse the performance of queries across multiple databases. While not providing a comprehensive guide, the tool highlighted potential issues, such as syntax errors, which could be investigated further,” Dhananjaya says.

IBRS notes that ManageEngine is more commonly known as an IT service management (ITSM) platform. However, it includes a less-known application performance management (APM) module. While this module did not provide as comprehensive analysis as competitive products, it was sufficient to identify the application’s issues that could be investigated further. Too often, when organisations realise they have an ICT problem, there is a tendency to look for a new, specialised tool to help fix the problem. It is far wiser to look at existing platforms and investments and leverage ‘more than good enough’ capabilities.  

These efforts paid off, as the system performance improved dramatically, and the speed of operations increased significantly. One of the key improvements was the implementation of a scheduling system that allowed reports to be generated overnight.

“This change drastically reduced the number of users accessing the servers simultaneously, improving overall efficiency by approximately 60 per cent. Teams that previously spent excessive time generating reports could now focus on their core responsibilities,” he says.

The company also made enhancements using an extract, transform and load (ETL) tool, alongside Tableau, for analytics. By pushing many reporting functions to the ETL tool, Dhananjaya was able to create dashboards for various departments, including the factory, and provide data in Excel formats tailored to different teams.

This approach not only reduced the load on their systems but also allowed them to optimise SAP licence usage by allocating licences more effectively based on user needs.

Furthermore, the company initiated a segregation of duties (SoD) project to enhance accountability and traceability within their transactions, further streamlining operations.

“The efficiency gains were substantial, with one application originally running on six cores being optimised to operate effectively on just four cores, maintaining a load of around 60-67 per cent. This optimisation translated into significant cost savings by reducing the hardware footprint and avoiding unnecessary investments in additional resources,” Dhananjaya says.

The optimisation efforts also fostered a culture of ownership among the teams, who now receive weekly performance reports from system managers. This transparency has helped eliminate the firefighting mentality and encouraged proactive management of system performance. The company has also established a process for evaluating new applications, ensuring that teams provide data growth projections to guide hardware investments appropriately.

According to IBRS, web application performance is impacted by physical infrastructure, non-scalable application design, technical debt, software, specialised services and WAN optimisation. This begs the question, who is actually responsible for a web application’s performance? IBRS recommends that a single person, or team, be responsible for end-to-end web application performance, with direct governance of the physical infrastructure, software and services needed.

IBRS also advocates identifying which IT groups are involved in performance issues for which web application projects. Consider how much overlap is there between groups, and where specific skills are located. Using this information, create a team to oversee web application performance as an end-to-end objective across the enterprise.

Transitioning to VMs

International Tractors faced challenges in managing its IT infrastructure effectively, often responding to increasing workloads by investing in additional hardware without properly assessing actual needs. This reactive approach led to overspending and potential compliance issues due to mismanaged licences.

To address these challenges, Dhananjaya transitioned the company to a fully virtualised infrastructure, allowing for more flexible resource allocation.

He says, “By moving all servers to VMs, the company gained the ability to quickly allocate additional computing resources  in case of application spikes. With the infrastructure now fully virtualised, the resources can be allocated dynamically based on demand, without the need for further hardware investments.”

Virtualisation also improved the segregation of workloads across its systems.

“For instance, the core dealer management system (DMS) was previously running alongside a non-critical database, consuming CPU resources and affecting performance. By isolating these applications on VMs, we could easily replicate the environments and ensure that critical applications were not impacted by non-critical ones. Each application now operates on dedicated IP addresses, enabling clear accountability and management,” adds Dhananjaya.

IBRS also notes that virtualisation initiatives go hand-in-hand with licensing consolidation. By implementing observability solutions, Dhananjaya ensured that International Tractors had the right number of licences, avoiding the common pitfall of underestimating requirements. 

“As a result of these efforts, we also optimised our Microsoft licensing strategy. Initially considering a significant investment in double digit core licences, the company found itself with four extra core licences available after the optimisation exercise. This proactive approach has allowed us to avoid unnecessary expenditures while preparing for potential future workloads,” he says.

For organisations looking to consolidate computing infrastructure with virtualisation, care must be taken to adhere to the nuances of licensing on VMs and when running on cloud-based virtual environments.

Virtualisation is a well-established approach and for many organisations has been subsumed by cloud platforms (which are largely dependent on various forms of virtualisation). However, larger organisations are beginning to realise that some cloud workloads can be more economically run with highly virtualised on-premises data centres, or within managed services. Therefore, it is worth reviewing best practices for virtualised environments, including:

• Design VM deployments with high availability and disaster recovery in mind. 
• Leverage features like availability sets, fault domains, and geo-redundant storage.
• Set up monitoring and logging solutions to track the performance, health, and usage of your VMs, enabling proactive issue detection and resolution. 
• Leverage automation and scripting of VM setups to provide both efficiency and a standard approach to security. 
• Periodically review VM deployments, identify underutilised or over provisioned resources, and make adjustments as needed to optimise costs and performance.
• Implement elastic workload balancing, potentially with bursting workloads to cloud resources during seasonal or sporadic peak loads take care to fully understand the impact of virtualisation on core computing services, such as the server operating system, database servers, and core commercial applications.
• Consider management solutions that can support physical and virtual on-premises and cloud computing infrastructure with a ‘single pane of glass’ approach.
• Consider leveraging cloud virtualisation through hypervisors; this approach allows companies to achieve resource efficiency, scalability, isolation, and enhanced security.  It forms the backbone of modern cloud infrastructures, enabling businesses to optimise their IT resources and adapt quickly to changing demands.
• As a best practice, companies should consider containerisation as a technology for digital transformation, particularly when transitioning from legacy to modern applications based on specific requirements. Containerisation packages applications and their dependencies into lightweight, portable containers that can run consistently across various computing environments. Unlike traditional virtualisation, which virtualises entire operating systems, containerisation shares the host OS kernel while isolating application processes.