Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 4 Next »

BUSINESS DRIVER

Executive Summary - Data driven placement and monitoring of workloads and infrastructure require AI based big data analytics. To support multiple edges, federated learning, it is required to deploy anlalytics packages at multiple locations.  This project intends to simplify deployment and automation of big data framework at multiple locations, thereby reducing the analytics deployment from weeks to hours. 

Business Impact - Simplify operations and separate out analytics framework from analytics applications, which allows multiple data scientist organization to deploy training/predictions applications without worrying about on how analytics framework is deployed and managed.

Business Markets - Applicable to physical and virtual network functions deployed in large operational networks - cellular service (4G/5G), cloud service and data center networks.  

Funding/Financial Impacts - Potential of significantly reducing the CAPEX

Organization Mgmt, Sales Strategies - There is no additional organizational management or sales strategies for this use case outside of a service providers "normal" ONAP deployment and its attendant organizational resources from a service provider. 

Technical Debt 

Features that would make it in for R4

  • Container images and Helm charts for 
    • Collection and Distribution stack (CollectD, Node-Exporter, cAdvisor, Prometheus, CollectD operator)
    • Data lake (HDFS and M3DB, M3DB operator, HDFS writer, Operator for HDFS writer and prometheus related remote_read and write services)
    • Training (Spark with scikit-learn, math libraries, MLLib etc...,  Horovod,  spark-k8s-operator and possibly Horovod operator AND mode optimizer & storage service)
    • Inferencing (Tensorflow-Serving)
    • Messaging (Kafka-broker, zookeeper, Kafka operator)
    • Model management (Minio and Minio operator)
    • Sample applications
  • Placement of stack in various locations and configuration of them to work together.
  • Features that may be postponed to R5
    • Various operators

Release 6 goals:

  • Performance features 
  • Security (ingress, egress, inter-component,  RBAC for each package)
  • Way to customize  packages
  • Deployment using Multi-Cluster scheduler
  • GUI/CLI based configuration of stack
  • Monitoring of stacks - Ensure that each stack component has prometheus target.
  • Some real AI based analytics



  • No labels