Ultra fast queries leveraging AVX-512 vector execution.
AVX-512 vector units have been shipping on Intel processors since 2018. Sneller translates SQL directly to AVX-512 assembly using the Sneller SPRINT engine. This enables 16x more processing power with the same silicon, so query latency is much lower.
Sneller Serverless Cluster
Serverless requires that each node is completely stateless. This allows compute resources to be added and removed with minimal overhead. Sneller has implemented very efficient and high throughput transfers from object storage and makes extensive use of memory caching.
Bringing It All Together
Data is ingested directly into your S3 buckets. The Sneller Query Cloud instantiates compute threads based on incoming Kibana or SQL queries. Sneller reads the appropriate data from S3 to process the queries.