INTRODUCING THE GREENPLUM MODULAR DATA COMPUTING APPLIANCE The industry's first complete big data analytics appliance

Extreme Performance with Elastic Scalability

Greenplum® Database and Greenplum Hadoop are at the heart of the Greenplum Data Computing Appliance (DCA), offering both structured and unstructured data analytics. They offer shared-nothing, massively parallel processing (MPP) architecture optimized for BI and analytical processing. Both use MPP to achieve near-linear scalability, with performance and capacity increasing linearly with the number of nodes.

Move Computation To The Data

Another core principle of Greenplum software is to collocate processing and data. Doing so enables DCAs to process database queries and MapReduce tasks in a fully parallel manner, using multiple, node-by-node storage connections simultaneously to acquire data for analysis. Collocation of computation and storage yields the greatest possible performance for analytical processing, while alleviating computation and network bandwidth burdens on other data center infrastructures.

Modular Design and Integration Facilitates Data Co-Processing

Growth of unstructured data leads many users to seek a combined solution for processing structured and unstructured data. In this scenario, unstructured data is processed in Hadoop, and resulting extracted structured data is merged with database data for further analysis. The Greenplum DCA provides fast parallel connections between Database and Hadoop Modules, enabling co-processing of varied data types. Greenplum Database transparently and efficiently queries data in both the database and Hadoop without duplication.

Enterprise High Availability

The Greenplum DCA meets the reliability requirements of the most mission-critical enterprises by delivering multi-level, self-healing fault tolerance, which includes automated failover, fully online self-healing resynchronization, and multiple levels of redundancy and integrity checking. Data availability employs hardware RAID protection with hot spare drives on standby for minimum recovery time. This ensures no data loss and minimal performance reduction during automated RAID disk rebuilds.

Reliable Backup and Increased Availability

Enhanced availability and disaster recovery are available for Greenplum Database Modules through EMC Symmetrix VMAX. VMAX snapshots and Symmetrix Remote Data Facility (SRDF) replicate data from the Greenplum Database to distant data centers for archival and maintenance of another DCA in warm standby.

Multiple HA configurations are available for Hadoop users. Users of Greenplum HD Modules enable Hadoop software to maintain a configurable number of data replicas, distributed throughout the Hadoop cluster. Users of Greenplum’s HD Compute Modules (which rely on EMC Isilon for HDFS storage) can leverage Isilon’s redundant storage and distant replication to maintain archives as well as a warm standby replica of the primary Hadoop cluster and its data.

Proactive EMC One Support Structure

Customer Support Services provide the resources and services to quickly and proactively resolve solution-related issues and questions, thereby assuring business continuity and a highly available data environment. EMC’s global maintenance and support is available around-the-clock 24x7 via live chat, online service request management, live telephone support, and onsite support through the industry’s leading global field service organization.

In addition, the Greenplum DCAs support EMC Secure Remote Support (dial-home). Through this feature, the appliance provides around-the-clock remote and pre-emptive troubleshooting by automatically alerting the EMC Support Center of critical hardware and software errors for Greenplum Database, HD and DIA Modules. The EMC Support Center can then remotely diagnose issues and dispatch replacement hardware and support technicians as required.

DCA Cluster Configuration

Greenplum DCAs can begin as small as ¼ rack and be expanded to as many as 12 racks, combining one or multiple module types. All modules within the DCA share a redundant 10GB Ethernet interconnect for maximum co-processing performance and availability. Administration is conducted over a separate 1GB Ethernet interconnect to support single-point administration of the entire DCA.

DCAs with Greenplum Database Modules begin with initial host processor nodes, a primary and a standby and a minimum of one ¼ rack Database Module in the first DCA rack. From the initial configuration, up to 48 total database modules can be added. As additional modules are added to the database, they add segment processor nodes to the MPP database cluster, to a maximum of 48 modules or 12 total racks.

Greenplum HD Modules work somewhat differently, not requiring a separate host processor pair. Instead they dedicate the first Hadoop Module to support JobTracker and NameNode functions. A minimum Hadoop configuration therefore begins with two modules, scaling to a maximum of 12 racks, just as does Greenplum Database.

Whether your ideal DCA configuration is pure database, pure Hadoop, or more commonly, a mixed cluster of module types, required network switching is included in each rack as needed, to maximizes performance while maintaining redundancy, across a single-rack or 12 racks.

Greenplum DCA Technologies