A company’s applications use Apache Hadoop and Apache Spark on-premises. The infrastructure is not scalable and is complex to manage. A solutions architect must design a scalable solution that reduces operational complexity while keeping data processing on-premises. Which solution meets these requirements?

The correct solution is to migrate the Apache Hadoop and Apache Spark applications to Amazon EMR clusters on AWS Outposts. This meets the requirements because AWS Outposts extends AWS infrastructure, services, APIs, and tools to on-premises facilities, allowing the company to run EMR clusters locally. This reduces operational complexity by leveraging managed AWS services while keeping data processing on-premises, as explicitly required. Using AWS Site-to-Site VPN to access on-premises HDFS and process with EMR in the cloud would introduce latency and doesn't keep processing on-premises. AWS DataSync is for data transfer, not for running applications or processing data on-premises. Using an AWS Snowball device to migrate data to S3 would move the data processing to the AWS cloud, which contradicts the requirement to keep data processing on-premises.