Jan 8, 2014

BoT, BoX or BoH (BW-on-HANA)-Version 2

Question: What does Radio VOE (Voice of Experience) recommend about migrating  SAP-BW to

[1] Teradata;

[2] Exadata/Exalytics; or
[3] HANA

Ask 1: What is the major difference between Oracle Exadata+Exalytics and HANA
Oracle Exadata is a standard RDBMS row storage data base. It uses the legacy methodology for modeling that includes. In order to run queries as fast as BW Accelerator it needs Exalytics (Which is exactly like the BWA)
1. Extraction to a PSA
2. Storage to an DW layer: ODS and Cubes
3. Indexing
4. Aggregation
5. Caching summaries
6. Query calculation in a separate server
7. Query Optimization in Exalytics (separate Server)
8 Query Results and rendering
All this creates multiple copies of data, is infra heavy and creates a latency between Data creation and Analytics (Decision Systems)
  
Visualization  1: Attributes for customers with SAP applications as source system for 50%+ of critical data




This above view represents a visual radar as to the capabilities of BoT, BoX and BoH











Visualization  2: Same information different visual
 
This visualization is a tabular view of Teradata, Exadata+Exalytics and HANA







      1.        BW-on-Teradata (BoT)
In the early 2000’s SAP partnered with Teradata as an alternative database for BW. To cut the story short globally there is currently only one customer with BW on Teradata and probably will always remain as the only customer to have ever taken a very short trip on a BoT (BW on Teradata). You don’t want to go there or ever take this path.. end of story
So that should lay to rest any endeavor for a BoT migration
PS the customer is currently seriously considering migrating their BoT to BoH in 2013..

Recommendation: Keep away from even trying a POC for a BW on Teradata.
     2.       BoX (BW-on-Exadata) vs. BoH (BW-on-HANA):-
F     For BW-on-Exadata the foundational advice would be similar to the one for BW-on-TD. If 99% of the planet is going on one path don't try to reinvent another one on your own. There is inherent safety in numbers. Radio VOE has validated that the strategic advantage of following the herd far outweighs the adventurist assumptions & costs of being a loner, ie. Having the self-cannibalistic honor of being the 5:3000 implementations on a unique track does carry its own risks and disadvantages. The key decision is not only about tactical cost but about strategic Business Value.
       Keep this as your 'center-of-the-universe' decision factor of your final decision.
Recommendation: Let the strategic benefit and the process drive your decision, not some business emergency creating unrealizable timelines due to inefficiencies of the past, drive your process.

Visual 3- An Oracle chart from Oracle Site on Exadata

This screen shot is from an oracle website for Exadata. It is important to note that at best Oracle can take a 50 minute Query down just below 7 minutes, another 40 minutes query down to 3 minutes and 40 seconds. The best they can report is 40 seconds to 4 seconds










Now let us dive together and look at what lies under the cover

Attributes
BoX (BW on Exadata)
BoH (BW-on-HANA)
What it really is
Database
Business Application platform
Does it need anything else
·  Mostly will need Times 10 for data load
·  Exalytics for OLAP acceleration
Nothing else
Core Strength
Write, i.e. getting data in
Read, i.e. getting data out
What you get on the BW side
Database
·  Business Application Platform
·  Over 300 RDS solutions
·  Automated Database slimming applications
·  Predictive analytics platform
·  Prescriptive analytics capabilities
What you get with SoH
·  ECC (ERP)
·  CRM
·  SCM
Database
·  Over 900 prebuilt views for real-time operational views
·  Prebuilt Operational analytics that come free with SoH
·  Access to state-of-art analytics and informatics
·  Predictive analytics platform and algorithms
Database Type
·  Exadata is standard RDBMS
·   Exalytics is Columnar for OLAP
Columnar and In-memory
Apples to Apples
Exadata + Exalytics (appliance)
HANA (appliance)
Pricing
·  Exadata for the db write
·  Plus Exalytics, for the read (analytics)
·  A license for every non PRD system including DR
·  Data base license, Plus Analytics engine
·  OLAP engine
·  Calculation engines and other applications
·  Check the total and final cost of ownership
All in one HANA platform and one price for PRD HANA. All other systems in the landscape come free

Master Node RAM availability
NA
1TB currently, up to 8TB by Q2 of ’14 and up to 16TB by Q4 of ‘14
Similarities
Exadata is an appliance
BW on HANA is an appliance
ACID compliance
NA
Fully Compliant
Need for MOLAP, Aggregates, Caches
Yes
No
Very High Strength
·  Structured data - Oracle ERP
·  No seamless SAP integration
·  No seamless comprehension to SAP  config changes
·  Structured Data- integration to SAP Applications
·  100% seamless integration to SAP Master and transactional configurations
·  Seamless merging of multiple master data and source system deviances – on the fly in memory mergers

Medium Strength
Seamless
Seamless Integration to semi Structured data
Weakness
Unstructured pure Big Data
Partnered with Hadoop
·  Unstructured pure Big Data
·  Partnered with Hadoop providers ; Funding Hadoop startups
Competition
Box compete with Teradata
HANA has no apparent competition for SAP customers.
No of Customers
NA
2400+ to date
Additional Needs
Will need Exalytics to provide accelerated queries
Does not need any additional accelerators
HW selection
·  Only from Oracle
·  Monopoly in HW development, support and services. No competitive forces encouraged.
Multiple global HW partners approved by SAP . This leads to customer advantage as it leverages automatic market driven  competitiveness and development between HW partners
Who are the top HW Suppliers
Oracle, Oracle and Oracle
HP is currently reported to be in the leadership position for HANA appliance installations followed by IBM

Strategic Impact
Does not align with SAP Suites strategy
Aligns to the HANA One-database strategy where customers get actual gains of HANA. Zero latency between OLTP and OLAP
How does it transform
·  Disc to server and back.
·  Data is passed from server to server till results are compiled. There is a need to break up responses
·  Database
·  All the calculations are done in memory in the db itself
·  No need to break up responses
How does it perform
·  Slower than Teradata
·  Oracle Benchmark: 10 bill records under 10 secs
·  Fastest on the planet
·  My experience 45 billion records under 10sec
·  SAP Benchmark- 240 billion records under 1sec
Maturity of the technology
 
Is equivalent to BW Accelerator (thus 2006)
Undisputed global leader
Point of Transformation
OLAP Layer – 90+%
This is the worst point of transformation
Better is in the DW layer
Optimal point of transformation
= database layer
 
 
 

How does it work
·    1.  Database server to
·    2.  Compute server to
·    3.  Storage cells to Exadata cells
All in-memory
Data Loading
·  ETL files in DBFS
·  Oracle golden Gate
·  Streams
·  3rd Party
·  Data integration tools
·  SLT (Real-time from ECC
·  BODS for batch loads
·  DXC direct BW
*** License
Every system in the landscape is subject to a SEPARATE license
1.        DEV, Test, QA, and PROD
2.        HA, RD and Training
3.        Other, i.e. Projects landscapes, etc..
Only PROD is licensed
All other systems and Free (not licensed)
What you get
Traditional DB accelerator
1.       Traditional ‘Wait and Debate’ engine
2.       Database
3.       Transaction Engine
Everything else supposedly are separate tools and licenses
Breakthrough HANA Platoform
1.        Real-time ‘Business NOW’ engine
2.        Database
3.        Transaction Engine
4.        Dimensional Engine
5.        Analytics Engine
6.        Predictive Engine
7.        Text Engine

Modeling methodology
Traditional- 10+ steps
1.      ETL
2.      DW
3.      ODS
4.      Cube
5.      Caching
6.      Indexing
7.      Aggregation
8.      Query Calculation Engine
9.      Query Transform
10.  Query Rendering
= RESULTS
Breakthrough 4-steps
1.      ELT/ ETL
2.      DSO
3.      In-Memory Analytics
4.      Query Rendering
= RESULTS
Impact of modeling method
·         Latency between operational and decision data
·         Infrastructure intensive
·         Expensive by system resources
·         Creates duplicates and is slow
Up to 1,000 times faster
No indexing or aggregates
Calculation / Transformation methodology
Traditional:
·         Most at OLAP (Query) layer – extremely inefficient
·         Rest in DW layer                    - Very Inefficient
Breakthrough calculation
·         Direct database transformations
·         Real-time db calculations
·         No mid-tier bottlenecks
Data-Center Impacts
·         Typically RAC (extra Costs)
·         Storage uses ASM failure groups
·         Box is a single point of failure
·         DR is a full license
Uses MPP automatic failover
Disaster recovery
·         License: full license for HCC, and even remote mirroring
·         2nd Exadata server/ non-Exadata server
·         Decompression to standard Oracle server possible with considerable performance loss
License: No extra Charge
Provided by HANA vendors

For SAP customers
 You just get another database. That too mostly RDBMS
All new functionalities only being developed on HANA
1.        BW seamless upgrade & migration in one step
2.        Over 900 views for ECC
3.        New RDS for CRM
4.        New RDS for SCM
5.        Industry specific solutions

Just FYI in 2013 SAP, HP BMSoft and Red-Hat broke the Guinness World record for the largest data load at 34.3TB in one hour
The certificate is attached to the left..


Happy reading to this second print as the first one had got damaged in translation and lost most of its table properties, it is now back but I will keep updating this one as more information becomes available



1 comment:

  1. Hi,
    Thanks for providing nice excellent information about sap hana modules
    I was very interested to read this information of hana and visit
    sap hana online training

    ReplyDelete