The Center for Data Innovation spoke with Leandro Vaz, co-founder of XONAI. XONAI provides cost-effective data analytics solutions that integrate into companies’ existing data platforms and cloud computing infrastructure.
Kir Nuthi: How does XONAI accelerate data processing through what it calls a “universal compute fabric”?
Leandro Vaz: The established paradigm in compute-intensive software (such as big data and AI) is to offload some of the execution to application-specific chips. Big data analytics in the larger scope, however, runs on an interconnect of extensive data pipelines, from stream and batch data ingestion capturing insights to all forms of data preparation needed to train large machine learning models to be deployed on data centers and down to mobile and edge devices. The optimal solution should be able to seamlessly map individual data pipeline steps to the best suitable hardware that can run each optimally, an idea we baptized as “universal compute fabric.” We are leveraging novel compiler technologies to future-proof our solution and enable supporting existing and emerging accelerator hardware for data analytics.
XONAI embraces open-source software for data analytics and respects customers’ choice of applications and distributed computing infrastructure to deploy data pipelines at scale. Our proprietary technologies integrate and expand the data processing capabilities of software dominating the data analytics market, such as Apache Spark and Presto, and enable it to work with existing distributed computing infrastructure, either on the cloud or on-premises, and without requiring code or infrastructure changes.
Nuthi: Why did XONAI originally focus on general-purpose hardware for its data processing solutions?
Vaz: Our go-to-market solution targets data preparation stages prior to training machine learning models, which typically run on general-purpose hardware available on the cloud, such as Intel and Amazon Graviton processors.
A significant fraction of the data preparation market can be characterized by Apache Spark running on Hadoop or Kubernetes clusters on data platforms such as Amazon EMR and Google Dataproc. While XONAI supports these markets that largely run on general-purpose hardware, our technology is ultimately designed to extend to larger data analytics use cases, particularly in the scope of machine learning.
What we have already built as a pre-seed startup is no small feat, as our solution can already enable large data-driven organizations to shift focus from manual infrastructure optimization to value-adding projects.
Nuthi: How does XONAI slim down costs for cloud infrastructure compared to single-vendor data platforms?
Vaz: The majority of solutions on the market require customers to migrate their data and workflow to a closed-managed platform with the promise of increased productivity. However, they also charge significantly more than baseline cloud costs and, in addition, lock users into a proprietary workflow that may not always be ideal as organizations scale, and this is often only realized post-migration.
XONAI’s products benefit data-driven businesses by lowering cloud costs without migrating off of their existing infrastructure and choice of software. Our platform is priced to capture the economic value of faster data processing delivered in customers’ chosen data platform, which can then run on cheaper machines more optimally or reduce time spent processing data on the same machines.
Nuthi: What are the regulatory issues XONAI has faced and predicts to face as it continues to refine its technology and expand globally?
Vaz: As we grow our operations, customer data protection is a top priority. Like any other company processing data at scale, demonstrating data security and compliance is crucial, as well as meeting any local, national, and international regulatory requirements for data protection. We are working towards obtaining SOC 2 compliance and ISO 27001 certification in order to ensure we have the proper business controls and management processes in place.
Nuthi: You’ve previously expanded your technology to integrate into cloud services like Amazon EMR and Google Cloud Dataproc. What’s next?
Vaz: Our platform is compatible with cloud platforms such as Amazon EMR and Google Cloud Dataproc, and we will continue to broaden software and hardware support to enter new markets. Supporting new hardware, however, without relieving the customer from the burden of manually and optimally mapping it to potentially thousands of data pipelines, would be like building a car without an engine or steering wheel.
XONAI wants to realize the next-generation platform for the cloud, which uniquely combines the serverless computing model with automated hardware selection for data analytics—in other words, transparently selecting the best hardware to run each individual data pipeline in order to meet the best performance on any type of distributed computing infrastructure.