At a high level Opni is made up of a number of microservices communicating via NATS. Logs are persisted in Opendistro Elasticsearch and AI assets are stored in S3.
Opni ingests logs through a http receiver for logs shipped from fluentd.
GPU service (optional)¶
Opni requires an NVIDIA GPU - we recommend at least a k80 - to learn and inference on your cluster's workload logs.