About
Provides an overview of HPE Ezmeral Unified Analytics Software.
HPE Ezmeral Unified Analytics Software is usage-based Software-as-a-Service (SaaS) that fully manages, supports, and maintains hybrid and multi-cloud modern analytical workloads through open-source tools. HPE Ezmeral Unified Analytics Software separates compute and storage for flexible, cost-efficient scalability to securely access data stored in multiple data platforms through a simple user interface, which is easily installed and deployed on private, public, and on-premises infrastructure.
Features and Functionality
- Access data anywhere and manage it in one place
- Connect bidirectionally to multiple data platforms and join data to create a federated data mesh that you manage in one place. Includes authentication, authorization, logging, metrics collection, and monitoring.
- Robust, integrated storage layer
- Includes an integrated, scalable data fabric storage layer with data-mesh like capabilities as the ephemeral storage for all types of data, including structured and unstructured data, files, objects, and streams.
- Analytical workloads
- Support for the most common enterprise analytics use cases ranging from traditional BI/Reporting (via PrestoDB and SparkSQL interfaces) to emerging workloads, such as exploratory data science, real-time analytics, and machine learning workflows.
- Self-service data access
- All users, including administrators, data engineers, data analysts, and data scientists can directly access data from HPE Ezmeral Unified Analytics Software.
- Built-in access to BI dashboards and data science tools
- Includes built-in BI dashboards for analytics and operational reporting, Also includes web-based notebook interfaces, such as Jupyter Lab and Visual Studio, for data science workflows (model training and serving frameworks).
- Built-in SSO
- Supports single sign-on experience; users sign in to access HPE Ezmeral Unified Analytics Software and compute components integrate with the storage platform infrastructure to pass the identity of each user.
- Performance
- Distributed, in-memory caching ( explicit) that accelerates federated queries on commonly used datasets.
Compute Components
The compute components included in HPE Ezmeral Unified Analytics Software enable users to get up and running in minutes. Components connect to each other at start-up and use pre-defined storage areas in the built-in data fabric. When applicable, compute components can automatically take advantage of GPUs.
- Spark
- Spark is a primary engine for data analytics tasks.
- EzPresto
- EzPresto is a distributed SQL query engine with a built-in query federation capability (distributed in-memory caching and pushdown optimizations) for fast analytic queries on data of any size.
- Kubeflow
- Kubeflow as an ML framework focused on model training that includes Notebooks, Pipelines (Airflow), Experiments, Kserve, and various distributed training operators.
- Airflow
- Airflow for data engineering and task automation.
- Notebooks
- Jupyter notebooks for performing varied data science tasks, such as cleaning data, labeling features, testing toy models, and launching distributed training models.
- Dashboard Frameworks
- Dashboard frameworks for building data models and visualizations.
Workflows and Pipelines
HPE Ezmeral Unified Analytics Software provides simplified workflows and pipelines for data engineers, data analysts, and data scientists to solve complex problems.
The following image shows some of the supported workflows and pipelines:
