Follow Datanami:
May 5, 2015

HP Targets Big Data Workloads With New Servers

Hewlett-Packard said it is targeting “data-intensive workloads” with a series of Compute platform servers and an accompanying big data reference architecture aimed at leveraging the data torrent from connected “things.”

HP said Tuesday (May 4) its new batch of ProLiant and Apollo servers were fine-tuned with a big data reference architecture to zero in on data-heavy workloads like Hadoop that increasingly dominate datacenters. The platforms also reflect the emergence of new server technologies and architectures that incorporate open source data analytics and in-memory databases designed to connect the dots and derive value from big data.

HP said the big data servers are designed to boost storage capacity, including block and file storage, while providing unstructured data and real-time analytics along with traditional relational database capabilities.

At the high end, the ProLiant DL580 Gen9 server targets data-intensive workloads such as in-memory platforms and structured databases like Microsoft SQL Server 2014. The Gen9 server is integrated with HP’s ConvergedSystem 500 to boost processing power, memory size and I/O capability to process larger data volumes, HP said.

A separate ProLiant BL660c Gen9 model is a four-socket Compute server blade intended to transform legacy converged infrastructure to increase the speed while reducing its datacenter footprint.

The Apollo 2000 and 4000 servers are intended as a bridge to scale-out infrastructure and, in the case of the 4000 model, a three-server family that targets workloads like Hadoop and big data analytics along with object storage. HP said it would offer a range of capacity density and performance scaling choices with one and three nodes. The ratio of compute to storage can be changed to match workload requirements, the company added.

The server platforms are supported by HP’s Haven Big Data platform that includes its Vertica system along with HP HyperScale big data ecosystem partners. They include Cleversafe, Cloudera, Hortonworks, Microsoft and Scality. They also support open source projects like Ceph, OpenStack Swift and Apache Hadoop, HP said.

When combined with HP’s Moonshoot hyperscale systems, the Apollo 4000 servers can run the company’s big data reference architecture, which the company described as a new Hadoop infrastructure design. The framework is designed to deploy a standard Hadoop distribution in an “asymmetric fashion” intended to allow customers to scale storage and compute independently. The result, HP claims, is up to twice the performance in half the space.

(In a footnote, the company noted that these performance gains “depend on workload as not all big Data workloads are the same.” Further, it noted, “traditional MapReduce workloads had 16 percent better price/performance, with double the density and half the watts. HBase had twice the performance for an equivalently-priced system.” The results were based on internal company testing in April 2015, HP added.)

HP said the Apollo 2000, Big Data Reference Architecture, HP Integrity Superdome X and HP ProLiant DL580 Gen9 are available now. The HP Apollo 4000 family, HP ProLiant DL560 Gen9 and HP ProLiant BL660c Gen9 are expected to be available in early June 2015.

Recent items:

In-Memory Computing Goes Open Source

In-Memory Computing Is the Key to Real-Time Analytics