hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jean-Daniel Cryans <jdcry...@apache.org>
Subject Re: data mining
Date Mon, 24 Oct 2011 20:32:35 GMT
On Mon, Oct 24, 2011 at 1:21 PM, Jignesh Patel <jigneshmpatel@gmail.com> wrote:
> J-D,
> Thanks for answering. Actually your suggestions gives me a good starting
> point.
> Yes our tables are relational but not in a relational form so that makes my
> life little bit easier.  However we will have constant insertion, updation
> and few additional complexities which I do like to avoid talking.
> Another question I have here, there are four choices I can make for the GUI
> based on my limited knowledge about hadoop and hbase.

You should probably read up a bit more then. May I suggest:

Tom White's "Hadoop: The Definitive Guide":


Lars George's "HBase: The Definitive Guide":

> 1. REST - not good for frequent transactions
> 2. Thrift
> 3. AVERO
> 4. Hadoop Map-reduce
> Not sure which one to opt?

The one that fits your needs.

> -Jignesh
> On Mon, Oct 24, 2011 at 2:49 PM, Jean-Daniel Cryans <jdcryans@apache.org>wrote:
>> That's a 250$/hr type of question, I don't think you'll get much help
>> here unless you have some more specific questions or someone feels
>> _really_ generous of their time.
>> My free tip is going to be that you should first do a POC that will
>> lay down the basis for your project. Putting tables into the HBase
>> model can be easy and hard, it really depend on what they look like.
>> If you don't have relations and multiple keys, it's as easy as putting
>> all columns for a table into a single column family where the
>> qualifier is the name of the SQL column. Then there's the question of
>> whether you just need to do a one-time insert or you need both DBs to
>> be in sync for some time as that's gonna require a lot more brain
>> juice!
>> Good luck,
>> J-D
>> On Mon, Oct 24, 2011 at 11:17 AM, Jignesh Patel <jigneshmpatel@gmail.com>
>> wrote:
>> > We have a requirement to build Hbase based system where we have to
>> architect
>> > to consume at least 3000-6000 tables.
>> > Has anybody done it. I am just wondering how to architect them from
>> > relational database to nosql database.
>> >

View raw message