Archived Content

The following content is from an older version of this website, and may not display correctly.

High performance computer maker SGI has created a big data number cruncher for its partner SAP - a single-node appliance that can process whole terabytes of information in-memory within the HANA system.

The invention was unveiled at the Sapphire Now conference in Orlando, Florida, on June 3 2014.

It uses a scale-up, single-node architecture with coherent shared memory, which SGI claims has not been achieved before.

The means that enterprise resource planning (ERP), data warehousing and other processor intensive applications can run on a single, in-memory system that can handle loads of up to 6TB.

This avoids the limits and complexity of clustered appliances, SGI said.

It means that data centers can perform analytics, transactions and processes in real time.

As a consequence, it is possible for banks, manufacturers and retailers to maker for faster and smarter decisions, according to SGI.

President and CEO of SGI Jorge Titinger said: “They’re striving for the performance of single node systems beyond the current threshold.”

The first release of the appliance will contain an eight-socket UV system using Intel Xeon E7 8890 v2 processors coupled with 6TB of shared memory.

This is designed to scale to 32 sockets and 24TB of shared memory as a single-node system.

Expansion is achieved by adding extra sockets, SGI claimed, so the complexity of clusters is eliminated.

This means that performance increases consistently and here’s no need to redistribute data or balance I/O.

Shared memory communication is achieved by SGI NUMAlink seven interconnects.

To protect data against memory power loss, data and log files will be stored on disk using dual NetApp E2700 RAID arrays.

“Our customers can now get the full power of HANA now,” Titinger said.