Rubrik Boosts AI-Enhanced Cyber Resilience with AMD

Rubrik Boosts AI-Enhanced Cyber Resilience with AMD

Rubrik, the Security and AI Operations company, boosted cybersecurity with AMD EPYC™ CPUs, gaining cost savings and AI efficiency for next-gen protection

Cybersecurity is increasingly important in today’s hyper-connected IT landscape. Keeping ahead of the threats, with the help of AI, demands more computing performance every year. Leading security and AI operations company Rubrik found that switching its cloud and on-premises customer workloads to AMD EPYC™ processors is delivering the performance and cost savings it needs for the next generation of protection.

“When we started Rubrik 11 years ago, data protection was very fragmented,” says Arvind Nithrakashyap, Co-founder and CTO of Rubrik. “Companies would buy backup software from one vendor, backup storage from another, then media service from a third, and put it all together.” This gave no guarantees of data recovery in the case of an issue. “We saw an opportunity to reimagine what is possible.” At the same time, cloud and virtualization were on the rise, while the type of threat was changing. The risks 15 years ago were human error, hardware failure, or natural disasters. “Today, cyberattacks are the biggest source of threats to company data.”

Rubrik realized that it could build a platform combining data and metadata together. “As part of protecting data, we can also build data security applications on top of that,” says Nithrakashyap. “We started by addressing virtualization. A lot of customers were running VMware applications in their data centers. We expanded to include other hypervisors and unstructured databases. We also saw a big transition to the cloud. Since we had built software that understood data and metadata in a single platform, it was easy for us to take that and apply it to their cloud and then SaaS workloads We built a complete cyber resilience platform.”

AI-empowered security growth

“We also saw an opportunity with AI,” says Nithrakashyap. “A lot of customers struggle with collecting all the data that's needed without leaving it vulnerable during the process. We have built a platform that understands this.”

All these workloads require high performance. “Our platform is compute-intensive,” says Nithrakashyap. “Whether it's just core data protection or security applications, we are doing a lot of operations. On top of that, for security reasons, we run our data through machine learning models that detect unexpected changes. We do operations both on premises and in the cloud. Having a compute platform with the right kind of performance, efficiency, and cost is critical.”

The growing importance of AI is further increasing the challenges for computing performance. “Even before the generative AI boom, our applications were compute-intensive,” says Nithrakashyap. “Just the cost of compression itself can be significant. But there's even more compute involved with machine learning to train the models and apply the models in runtime. Generative AI lets us train models at a much higher level. It's not just about finding certain regular expressions but understanding the nature of documents, leveraging a combination of large language models to do that in a cost-efficient fashion. Those are extremely compute-intensive operations.”

Rubrik’s relationship with AMD has been bilateral from the beginning. “There was mutual interest on both sides,” says Nithrakashyap. “AMD felt our platform could help AMD secure its data and achieve full cyber resilience. Conversely, our customers were looking for us to provide more cost-effective compute platforms, both in the cloud and in the data center. As we engaged with AMD, we understood that AMD EPYC processors could deliver higher efficiency, but at a low cost.” With the difficulty of getting access to GPUs, deploying AMD EPYC CPUs for AI could also be more cost effective.

High performance and cost benefits from AMD EPYC CPUs

“We have gained a big advantage by using AMD EPYC CPUs,” says Nithrakashyap. “With AMD we see better TCO across the board, whether we are hosting within our own backend, or running on our customers’ infrastructure. With the high performance of the AMD EPYC CPUs, the better efficiency provides considerable advantages for us and our customers.”

“Our internal benchmarks showed AMD EPYC processors outperforming competitors both in the cloud and the data center,” says Nithrakashyap. “We also measured running costs and saw considerable savings, both for our internal cloud and for our customers’ data centers. Migration was straightforward, especially in the cloud. We just powered up our software in another instance, and immediately we started seeing the benefits. On-premises, we had to qualify applications and hardware configurations, but it was straightforward.”

Rubrik has been able to achieve a lot, thanks to the switch to AMD EPYC processors. “In the last three years there's been a push on efficiency,” says Nithrakashyap. “That efficiency is easy to demonstrate when we use AMD processors, particularly with AI, where you can run workloads with regular CPUs rather than GPUs.”

AMD is playing a key role in making these new workloads more affordable, particularly when customers are wary of using LLMs hosted in the cloud. “Most customers don't want us sending their data to an external AI platform. But the cost of running that locally is very high. In our experience, running AI workloads on AMD processors delivers better efficiency, fast performance, and lower cost. That’s the advantage of using the latest AMD innovations.”

Towards full cyber resilience

“Companies want to be able to leverage AI but the costs are very high right now,” says Nithrakashyap. “If we can start doing that in a more efficient way, that will enable a lower cost deployment of language models or algorithms. If you want access to a GPU in the cloud, you must pay a premium. Being able to process AI without GPUs is where AMD EPYC CPUs can be particularly useful. We are in the early stages of what we can do with AI, and we will be needing more compute capabilities to be able to run these workloads.”

“The fact that AMD can provide the same level of performance with smaller form factors has been very useful in the data center,” says Nithrakashyap. “We are exploring every opportunity to standardize on AMD as the default processor vendor for our internal data center and cloud use cases. We recommend the same to our customers. Over 50 percent of our cloud offerings run on AMD today.” Rubrik now has 66,640 vCPUs powered by AMD EPYC processors. “We are also launching some new AMD instances to increase that. On-premises, AMD is our lead option for new server infrastructure.”

“We have three pillars on our platform,” concludes Nithrakashyap. “The first is data protection across enterprise, cloud, and SaaS. The second is Cyber Recovery, which enables customers to be able to be proactive against cyberattacks. The third is AI Acceleration. All of these are highly compute-intensive operations where cost and efficiency play a huge part. This is why we've decided to double down on our partnership with AMD, ensuring that our customers have full cyber resilience right across the board.”