SAP NetWeaver BI Accelerator (BIA)

SAP NetWeaver BI Accelerator (BIA)

SAP is introducing new approach to raise the performance of its BI platform. Foundation of this approach is search and classification engine (TREX) and special hardware where BI solution runs. The goal is to get significant analytical processing performance improvement in terms of query executing via:

• horizontal data partitioning – BIA is divided into partitions, typically one partition per CPU
• shared nothing parallelism – querying is divided into multiple steps, one per partition, each step is executed concurrently by separate processes, since each step is processing different amount of data in single CPU, steps are performed independently without interfering to each other – they share nothing
• vertical decomposition/column based storage – BIA stores data separated by columns, instead of keeping all columns and rows all together
• data compression – comprises with column based storage with dictionary based algorithm that represents dictionary index as small as possible number of bits
• in-memory database processing – data retrieval for first time caches infocube data and indexes of each partition into memory – so called “in-memory database”

BI Accelerator runs only on SAP NetWeaver 2004s and you need to have hardware infrastructure based on Intel’s Extended Memory 64 Technology (EM64T) which adds 64bits capability to Xeon family of processors. Currently are HP (Blade Servers) and IBM (eServers) supported hardware vendors for BIA.

What you will get when you would buy it? BIA is delivered as appliance which runs BIA functionality with Blade based server equipped with Intel processor and build-in storage. Appliance is connected into existing SAP NW BI 7.0 (NW 2004s or higher). BIA uses an RFC connection to BI backend which is to be accelerated.

For more information about BIA see on SDN.