HPE has updated its HPE Ezmaral Software platform to make ML and AI projects easier to run. Think of open-source tooling, integration of object and streaming data sources and more interoperability with multiple cloud environments.
The latest version of the HPE Ezmaral Software platform for data analytics should end the chaos that all analytics tools and workloads entail around performance, cost and compliance requirements. Issues related to data control, potential vendor and data lock-in, and performance around public clouds are also addressed, as well as issues associated with home-built open-source tools.
In the new version, the HPE Ezmeral Data Fabric Software provides seamless access to multiple data sources and formats. Think of files, objects, tables and data streams. Users get, among other things, highly granular control tools and automated policy management for workload optimization. This based on performance, data location, sovereignty, costs and compliance requirements. This software is now available as a SaaS solution for the first time.
The HPE Ezmeral Unified Analytics Software suite offers multiple “as-a-service” open-source tools. There is the Apache Airflow workflow management platform, the Spark analytics framework, the Superset visualization platform, the Presto SQL distributed data store and the Ray distributed computing framework. The Feast store for ML, the Kubeflow platform for Kubernetes and the ML management platform MFlow are also available.
For training and deploying AI applications, this software offers various new connectors including Snowflake, MySQL, DeltaLake, Teradata and Oracle.
The new HPE Ezmaral Software platform is specifically designed for hybrid environments and can be used in edge, colocation, on premises and public cloud environments. This should save customers the cost of ownership because it performs the analytics where both the data and computing power are located.
Read also: HPE brings GreenLake cloud experience to on-prem block and file storage