Query Performance Business Case

Query Performance Business Case

Goal
Predictable performance to Ad-Hoc Queries

Alternatives• HW Change: get more powerful HW, get additional HW, i.e. BI Accelerator, add memory, etc
• SW Change: change database, change DW software, Teradata, BI Accelerator, Neoview, Upgrade BI, etc
• Remodel: Single EDW Architecture, enhance Architecture, Modeling, Automated InfoCube modeling

Findings
Reported=by vendor; tested= real life ‘N’x= times faster

HW Change: can result in a max of 4x times reported improvement. BIA can result in tested 18-60x performance improvement. Add memory tested 3x to 5x improvement

SW Change: database change from oracle to DB2 reported a max of 18x but has other issues to consider; non SAP solutions can increase performance by 40x-60x but add integration costs; BI Accelerator tested at 18x-60x improvement; Upgrade could degrade performance but plan for 1x-2x improvement with functionality changes.

Remodel: Single EDW tested at 4x-8x in cross SBU reports; Enhance Architecture provided 2x to 5 x improvements; automated modeling findings 30x to 90x improvements

Collaboration: Automated InfoCube modeling and automated create InfoCube from DSO ++ in combination with BI Accelerator tested at 100x to 180x

the 3 layers of BIA Support

The BIA is sold as a standalone appliance to most customers.
15 days after installation we all realize it is more that that.

BIA Has three distinct layers that require support.

layer 1: BIA SW
layer 2: BIA APpliance
layer 3: BIA Application Layer

layer 1 is supportd by SAP America and includes the software installed on the blades in the BIA

layer 2 is supportd by your HW vendor and it includes the full support of the appliance other than layer 1 support, i.e. the BIA Software. This includes the HW and SW in the storage, the blades, the BIA Manager and other applications and HW that constitute the BIA

layer 3 is something we find out after go-live. The application layer consists of tasks that the customer support team has to undertake.
Selecting InfoCubes that need to go into the BIA; Creating Infocubes for reporting DSO's that are currently used for reporting (DSO/ODS creats a reporting drag on BIA performance and must be currently converted to an infoCube to enhance ROI); Indexing Infocubes for BIA, Data reconciliation between BIA and BI, Planning and forecasting of the BIA, Viweing the space utilization in BIA, projecting BIA Blade order status, Providing a monthly snapshot of BIA status, etc.
All tis needs a RACI, ownership and accountability.

Plan for it and especially paln for deploying the support for the application layer in the BIA.
Look at www.erada.com and the tool called BIA Workbench

Query Performance is it HW, SW or Design

Goal
Predictable performance to Ad-Hoc Queries
we do not compete with pre-run or cached reports

Alternatives
HW Change: get more powerful HW, get additional HW, i.e. BI Accelerator, add memory, etc
SW Change: change database, change DW software, Teradata, BI Accelerator, Neoview, Upgrade BI, etc
Design: Single EDW Architecture, enhance Architecture, Modeling, Automated InfoCube modeling

note: BIA comes both in HW and SW as it is an appliance consisting of both

Findings
Reported=by vendor; tested= real life
Nx= times faster

HW Change/ Performance improvement
Can result reported improvement 04x-5x
BIA can result in tested at 18-60x
Add memory tested 03x-05x

SW Change:

Oracle to DB2 reported a max of 18x-20x
but has other issues that must be considered
non SAP solutions 40x-60x
Oracle, Teradata, nateeza, but add integration costs
BI Accelerator tested at 18x-60x
BI UpgradeUpgrade plan for 01x-02xc
note: ould degrade performance with functionality changes.

Design / Performance
Multi > Single EDW tested at 04x-08x
in cross SBU reports;
Enhance Architecture 02x-05x
Automated modeling 30x-50x

Collaboration / Performance
Automated InfoCube modeling 30x-50x
with BIA tested 100x-180x

for more details revert to author