Oracle BI Architecture Masterclass

Yesterday I’m fortunate enough to attend Oracle BI Architecture masterclass for partners. The class largely covers the new Oracle Reference Architecture version which you can read in great detail from thispdf file (please note that this file is actually related to BI Architecture version 2, the version 3 is not released yet but they're enough to understand the architecture).

If you are familiar with the previous Oracle reference architecture, you will find that it has the similarity of the component namely Staging, Foundation and Access and Performance layers. However it now adopts the most talk about BI buzzword ‘Big Data’ into its component and Data Discovery/Knowledge Discover that supports agile BI which has become a norm this day has a special place in the architecture.

During the introduction, the instructors showing us an Interesting radar graph that shows how the big data is not a replacement of the old good RDBMS technology but a complement to each other. Here is the graph that also can be found on page 24 of the above pdf file:



Even Facebook the biggest Hadoop user realises this paradigm (http://www.informationweek.com/software/information-management/facebook-exec-databases-hadoop-belong-together/d/d-id/1112126?) and has used RDMS as their analytic technology alongside Hadoop. In line with the Oracle suggestion, the Facebook exec says that they keep the low-level, most granular, detailed data in Hadoop but move the transformed and aggregated data into RDBMS for slicing and dicing.

There are a couple of surprising suggestions that they mentioned for the implementation of the architecture. One revelation is the extraction of the sources data should be done as a whole and as quickly as possible (there are a hot and interesting discussion on this issue as one of the partner objected that this is not always possible). In doing so, the data warehouse is more agile and any future changes that required new data items from the sources do not need new extractions or requests from the source owners. The other thing that came out of the discussion is whether Master Data Management (MDM) should be kept in data warehouse environment (or be part of data warehouse project) or should be kept out of the data warehouse. It seems reasonably enough that the suggestion is to keep it out as the data warehouse is complex enough for another one to part of it.

Everyone in the room seems to see a rapid growth of interest and implementation of big data in the past few years. The instructors which have several years of working with the clients also confirm it and they seem very optimist of the advance of the Oracle technology to support big data solutions. They also think that they can give clearer guidance on how to answer the not so clear ‘big data ‘ requirement from clients.

A few interesting points that came out of the class:
  •           Try to have small number components as possible. The architecture is not rigid so be pragmatic.
  •           Try to have small number components as possible i.e. be pragmatic
  •           Don’t prevent for the users to access their data on every layer
  •           Don’t attempt to build MDM on the data warehouse
  •           Don’t wait to ingest data into Raw data layer
  •           Consider to implement in-memory  solution
  •           Data literate users should be involved early on by giving them access to raw data layer
  •           Get Finance analytics (EPM) fed back to source system
  •           Don’t be scare to use COTS BI to integrate with Oracle Reference Architecture as they normally fit to it


Comments

Post a Comment

Popular posts from this blog

Unit Test Framework for ETL (OWB and ODI)

Creating a good unit testing admittedly is as challenging as coding itself. It requires the same attention and time even though it's one of the most common part of the development life cycle that is not taken very seriously (alongside documentation). In my previous project on OWB, I spent a lot of time creating an automated unit tests using utPL/SQL test framework. The challenges are even greater as everything has to be coded and compiled as PL/SQL packages. The framework does not have UI tools that can be used to speed up the unit test creation. But it provides many assertions libraries that suffice to say it can handle most of the assertions conditions that I needed. It took almost the same amount of time to create a unit test for one OWB mapping. However I felt it was perhaps the most productive time on the entire development cycle. It cut out a lot of manual steps which are prone to errors. It helped to identify broken mappings which are caused by DDL changes. Last but ...
Read more

Oracle Real-Time Decision

I came across Oracle Real-Time Decision (RTD) only recently although this product has been part of Oracle BI product family since Sigma Dynamics acquisition in 2006. I was on the bench and skilled up my Oracle BI products after a year working in a project with no time to do any training. RTD is called by several name: 'a closed-loop recommendation engine', 'predictive analytic tool','real-time decision engine' and many more. I personally like the 'decision engine' name as it's actually a decision learning engine with low latency processing so it can make a decision in real-time. There are plenty of blogs about RTD. You will find that Rittman Mead blog has the most entries on the subject (as per other Oracle BI products - I have been told that they have a special relationship with Oracle that they can get the first hand experience of the products before Oracle released them to the public), they can be found here  and search for 'Oracle Real-Ti...
Read more

Pengalaman saya mengajar OBIEE

Dalam jenjang waktu setahun masa kerja saya di Hitachi Consulting, saya telah dua kali mengajar OBIEE di kelas. Training pertama diselenggarakan sewaktu bekerja di proyek untuk Anchor Trust (http://www.anchor.org.uk). Perusahaan charity (mungkin bentuknya seperti yayasan) yg menyediakan Home Care (pondok jompo) buat para manula ini bermaksud mengganti Business Performance Monitoring mereka dengan OBIEE. Saya bertindak sebagai OBIEE consultant yg bertugas untuk membangun dashboards, reports dan KPI. Sebagai bagian dari knowledge transfer saya diharuskan melakukan pelatihan penggunaan OBIEE secara umum dan cara memelihara BPM yg kita bangun. Jumlah peserta ada 6 orang. Tidak ada diantara mereka yg punya pengalaman menggunakan OBIEE, tetapi 3 diantara mereka adalah staff IT yg sangat 'techie' dan berpengalaman menggunakan dan memelihara system lama perusahaan ini. Mereka dapat dgn mudah menyerap segi-segi technical dari pelatihan seperti misalnya OBIEE architecture, RPD dan Web ...
Read more