By Mike.Hallett@Oracle-BI&EPM on Jan 21, 2014
From our BI partner Forum last week in London focusing on Big Data Analytics, I think the messages coming through from the discussions with partners and clients can be distilled into a few points:
1. Find Value = Organisations are more successful in finding value from potential Big Data Analytics projects when they combine data from “Hadoopable” sources with their traditional Oracle databases: for example merging social sentiment information with existing customer loyalty and product revenue data. Hadoop projects in isolation are struggling to complete the business case.
2. Oracle+Hadoop = Open source Hadoop and “R”-analytics are very complimentary to the Oracle stack, and indeed are greatly strengthened by Oracle’s software distribution and connectivity options – notably with OBI, Endeca, ODI, SQL, and the Oracle Big-data-Connectors for Hadoop.
3. Interconnect = Any Oracle partner can get going with this by downloading the software onto a relatively modest server, where they can interconnect the complete platform, and quickly get to grips with it: see below.
a. You can directly access data from Hadoop to analyse, visualise, and assess its’ value, and join it to Oracle DB data within OBI and Endeca and R.
b. If and as needed, data in Hadoop can be economically selected, processed, and enriched with familiar tools like ODI, and then moved with the Connectors into an Oracle database at very high speed transfer rates, within your data governance model (e.g. see the Oracle's Information Management Masterclass Tutorials).
4. Start Smaller = Of course, an Oracle Big Data Appliance plus the Exalytics and Exadata engineered systems are the ultimate solution for this, but all of the software platform from Oracle runs very well with Hadoop on smaller commodity servers. The choice of scale depends on your client’s needs – this is not a constraint on getting value from the integrated software stack. Projects can start relatively “small”; with one commentator suggesting the median deployment is around 10 Tb, and only very few are in the Petabyte range).
5. Enterprise-grade = For the successful Enterprise-grade projects, the Oracle tooling that connects with Hadoop and R provide the security, speed and reliability needed to de-risk the project and surface the value. These familiar tools (OBI, ODI, SQL, Big-data-Connectors) also minimise the skills gap and cost otherwise needed for extensive Java and MapReduce programming.
6. Hadoop is not Big Data = The key to value is using the right set of tools for the job: this may or may not include Hadoop.
This is a recent customer story where the full architecture is being put to broad use:
CaixaBank deploys new big data infrastructure on Oracle - Enables the Spanish Bank to Develop Better Customer Services, Integrating Data Created Across All Its Channels.
Interconnect the complete platform
Partners can assemble a complete platform from these VMs. These pull together a large set of tools, and so I recommend you put this on a server with 32Gb RAM. You can also get some hints @ Oracle BI Partners Get to Grips with Big Data Analytics
- Download OBI-Foundation SampleApp VM - including OBI, R, ODM, Endeca, Oracle-DB, Times Ten, and Essbase @ Oracle BI SampleApp V309 on OTN
- Download the Cloudera VM @ http://go.cloudera.com/vm-download
- Download Oracle Connectors for Hadoop @ Oracle Big Data Software Downloads
and White-paper @ “Information Management and Big Data Reference Architecture”
- Tutorials Introducing Big Data
- Oracle R Enterprise Tutorial Series
- Video Lecture Series Data Science Boot-Camp Tutorials