SAP this week not only announced the general availability of its namesake ERP application on the SAP HANA in-memory computing platform, but also revealed a new technology partnership with Hewlett-Packard under which it collaborated on the development of an x86-class server configured with 12 terabytes of memory.
Known as Project Kraken, the prototype system is based on 16 Intel Xeon E7 class processors (codenamed Ivy Bridge-EX) to make 240 CPU cores.
Speaking at the Sapphire Now 2013 conference, Prof. Hasso Plattner, chairman of the SAP supervisory board, says HANA is capable of scanning 3GB of data per second per core in the system it runs on. As such, Plattner says that SAP HANA represents a major advance in massive parallelization that allows applications to infinitely scale as new processors are added to the system.
In addition, SAP has rolled out an SAP HANA Deployment Shell that provides IT organizations with frameworks for optimizing application processing across multiple cores. Available on Github, the SAP HANA Deployment Shell is intended to make adopting parallelization constructs when developing applications.
Available on premise or in the cloud, SAP HANA leverages a columnar database designed to run across DRAM and solid-state storage (SSDs) in a way that delivers orders of magnitude of processing performance in a way that not only eliminates the need for batch processing, it allows analytic and transaction workloads to run simultaneously in real time.
At the same time, by taking advantage of more efficient approaches to SQL and stored procedures and eliminating reliance of disk-based storage, the total cost of managing the data center environment shrinks by as much as 30 percent, largely because of the application of data compression algorithms within the SAP HANA platform.
While HANA at its core is based on a database, the architecture of HANA provides support for different types of engines that can handle text analytics, spatial or geo-spatial data or even legacy relational database applications. The system makes use of a columnar database construct and a “Delta” data store to fetch different data types being processed by each individual engine via a layer of data virtualization software through which queries can be bi-directionally sent between HANA, traditional database systems and Hadoop.
SAP says it has signed up 1,500 customers to run SAP HANA, most of which are still in pilot. But as more HANA applications become more prevalent, Plattner says that HANA will become a platform for running applications both inside and out of the enterprise.