Salesforce uses AMD EPYC™ Processor-powered N2D machine types on Google Compute Engine to boost performance and reduce costs associated with its mission-critical CI/CD pipeline.

Customer relationship management (CRM) software is one of the most popular tools companies use to build strong relationships with existing and new clients. As more customer-related data becomes available to all organizations, solutions that can generate insights from the vast treasure trove of data can make the difference between thriving or struggling to stand out in today’s competitive markets.

Salesforce is one of the world’s leading providers of CRM solutions. Over the past two decades, the company has become ubiquitous across regions and industries. Today, Salesforce continues to push CRM innovation, constantly delivering new capabilities that enable organizations worldwide to optimize marketing, customer service, application development, and other vital business activities.

“Providing our scope of applications at scale requires tremendous IT infrastructure,” says Vidur Apparao, Senior Vice President of Engineering at Salesforce. “We began leveraging the public cloud to more efficiently manage workloads, improve our developer and infrastructure agility, and speed product time to market in new regions.”

Salesforce relies heavily on virtual machines (VMs) to support the continuous integration and continuous delivery (CI/CD) pipeline fueling its flagship CRM software, spinning up as many as 250,000 VMs daily. After running VMs on-prem for years, it recognized the opportunity to reduce costs and improve performance by running AMD EPYC™ processor-powered N2D machine types on Google Cloud.

Google Cloud results:

  • Supports scale of 250,000 VMs daily to power CI/CD pipeline
  • Enables engineering team to conduct up to one million tests regularly
  • Reduces overall costs associated with underlying VM infrastructure
  • Enhances global reach through regionally available cloud services

Supporting vital testing environments with no disruption

To provide people with optimized experiences when using Salesforce CRM tools, the company’s engineers constantly refine the software’s underlying code. With each change, Salesforce executes more than one million tests to validate the code’s quality and ensure development is progressing as planned. All of this is done through VMs.

VM utilization is also impacted by seasonal shifts, feature freeze deadlines, and more, meaning that Salesforce needs the ability to efficiently scale its number of VMs up and down as work requires.

“Google Cloud gives us the infrastructure, agility, and workload support to scale our VMs as needed,” says Vidur. “With Google Cloud, we can move easily from being in just one region to several—all now offering greater reliability in our test systems. Google Cloud meets us wherever we need to be.”

If issues arise with the infrastructure supporting VMs, code changes and testing would stop as thousands of engineers would be unable to work. Google Cloud provides a highly reliable infrastructure foundation that can scale to Salesforce’s needs and required speed of development. This keeps engineers, particularly those working within the CI/CD pipeline, focused on their work without disruptions.

“The systems supporting CI/CD are critical for us to be able to deliver exceptional experiences to Salesforce customers. They have to run reliably, quickly, and flexibly,” says Vidur. “Google Cloud gives us the agility to grow or shrink our infrastructure as needed, take advantage of higher performance, and ultimately provide ideal tooling to our developers to be even more agile and innovative.”

Google Cloud meets us wherever we need to be.

Vidur Apparao

Keeping costs in check

With its infrastructure scaling to support an increasing volume of work, Salesforce looked to reduce costs while avoiding performance impacts.

“We could not cut costs if it impacted performance because we couldn’t have tests running slower or possibly even failing,” says Vidur. “When Google Cloud and AMD came out with the N2D machine type, we saw an opportunity to get the best of both worlds – better performance at a lower cost.”

N2D general purpose VMs are built on 3rd Gen AMD EPYC™ Processors, provide up to 224 vCPUs and are the largest family of machine types on Google Compute Engine. The VMs are particularly useful for high-performance computing (HPC) workloads like those Salesforce relies on for its CI pipeline, as they combine higher platform memory bandwidth and lower costs than other machine types.

“Now that we have the N2D machine type, we’ve been able to quickly scale our VM workloads and will continue to do so,” says Vidur. “The cost benefits we realize allow us to test and support more workloads for new products and keep everything moving ahead within budget.”

Along with the lower costs, Salesforce is better positioned to respond to issues as they arise and provide highly responsive customer services while continuing to deliver CRM innovations.

Partnering to further improve the CI/CD pipeline

Salesforce wanted a cloud service provider that could come to the table as a partner rather than a vendor.

“We felt supported in our journey to Google Cloud thanks to the people guiding us through the process,” says Vidur. “Whether we have an architecture question or need to quickly scale up new VM instances by 10 percent, a simple phone call to Google Cloud gives us everything we need.”

Before transitioning its VM infrastructure to Google Cloud, Salesforce had to plan infrastructure increases or decreases months in advance. Today, changes can be made within minutes, giving Salesforce the flexibility to manage its CI/CD pipeline fluidly regardless of spikes in demand.

AMD, Salesforce, and Google Cloud worked collaboratively from the outset of the project, relying on the power of the N2D machine type to make the project work. The promise of the technology helped to achieve the results required while deepening the partnership between the three companies.

Thanks to the collaborative approach, thousands of engineers dedicated to the CI/CD pipeline are able to make as many changes as necessary to continue improving customer experiences with Salesforce CRM. The high-performance, constantly available, and regionally accessible solutions also improve time to feedback on bug fixes, new feature rollouts, and more. Faster feedback in testing translates to quicker speed to market.

Growing the benefits of CRM

Salesforce continues to expand to serve new customers worldwide and benefits from its strategic investment in cloud services and technology partnerships. The company is moving more workloads, especially those that have regional data residency requirements, to Google Cloud to provide its global engineering teams with low-latency access to a responsive, agile development infrastructure.

“Improving our overall information and data management is a key initiative at Salesforce,” says Vidur. “Google Cloud enables our development teams to work reliably in the public cloud in every region and at the scale, speed, and flexibility they need to deliver better customer experiences globally.”