Re: oracle and hadoop exploration

  • From: Paresh Yadav <yparesh@xxxxxxxxx>
  • To: igor.racic@xxxxxxxxx
  • Date: Fri, 3 Oct 2014 14:34:28 -0400

Hi Max,

Few ideas below and some slightly OT.

   - Business might not "get it" and appreciate the analytics done on AWR
   data. If possible ask this guys if they can point to any public anonymized
   sample data in consumer product goodsspace (I
   don't work for them :))
   - Connect Excel to your Hadoop cluster and use some out of the box chart
   types to display your results, business love charts/visual and if they are
   in Excel they will like it even more (it is very easy using ODBC setup,
   Google it)
   - Coming to AWR data, what questions you would want to ask of AWR data
   from multiple systems? Does it make sense to relate AWR data from multiple
   systems? Interesting perspective, here are some sample questions that can
   be asked but do they make sense? *Do the answers provide any useful
   information, value? You might want to ask Tanel Poder. *
      - Top 5 wait events across all systems
      - Top 5 wait events across all systems vs DB Time/Elapsed Time,
      across all systems
      - .....


On Fri, Oct 3, 2014 at 12:58 PM, Igor Racic <igor.racic@xxxxxxxxx> wrote:

> Hi,
> Haven't tried, but maybe running it in database can be an option ?
> Some info
> Regards,
> Igor
> 2014-10-03 16:35 GMT+02:00 max scalf <oracle.blog3@xxxxxxxxx>:
>> Hello list,
>> our DBA team(including myself) are exploring Hadoop.  Our
>> company(consumer product goods) is a pretty decent size company (revenues
>> about 18-20 billion) and we are a big SAP shop here with all big bells and
>> whiles (SAP Hana running for few modules) but we also want to explore
>> Hadoop as a new option.
>> Where we are struggling is, some of the business folks do want to start
>> putting some data in Hadoop echo system and see what value they get out of
>> it but DO NOT WANT TO PAY FOR POC.  So our department came up with a idea
>> of doing a small POC(maybe a 3-4 node cluster) with the data we have in
>> hand.  What i mean when i say "data we have in hand" is, awr performance
>> data, syslogs from unix* server, windows event log on windows server,
>> whatever a DBA can get there hands on(except the business data)...
>> So here i am reaching out to community and see if anyone else have a
>> similar problem and what they have done about it?  Maybe a good use case
>> mining AWR performance data(from 100's of oracle system) on hadoop or
>> something like that.
>> I understand hadoop is not the answer to replace your traditional RDBMS
>> but we are just trying to step into that big data world and do some POC, so
>> the business can also jump in.
>> Any thoughts or suggestion or good use cases ??

Other related posts: