Comments inline

2009/8/4 Onur AKTAS <[email protected]>:
>
> I'm confused about the when to use HBase. (Sure I have read everything and 
> every link on http://wiki.apache.org/hadoop/Hbase/FAQ)
> Its distributed (when used on Hadoop) and provides efficient database for 
> "VERY LARGE" amounts of datas, thats ok.
>
> But as I come from RDBMS world, i want to learn what I can do with HBase and 
> how it can affect the quality of my applications.
>
> For example, as I also see on "HBase Schema Design Case Studies" referenced 
> by FAQ, "Case 4: user-friends", there is something like
> user: id
>         info: name
>         info: sex
>         info: age
>         friend: user_id1
>         friend: user_id2
> etc..

It's possibly worth pointing out, that a userinfo table will quite
likely fit fine inside a relational db like postgres.

>
> 1- Let me understand the concept, Is there anyway to delete user's id from 
> all of his/her friends' friend column? There is no foreign key relation then 
> this is the thing that we must handle programmatically?
>

That's correct, there is no way to do this automatically, there are no
relations in HBase.

> 2- Is it also useful when the data is small but the calculations on it is 
> complex? For example, "calculating the degree of contacts" ?
>

HBase 0.20 handles small cells fine, eg. I've loaded a 700 million
cell table of integers. You can run map reduce jobs over the whole
table if you want to do processing.

> 3- Some high traffic applications like making Facebook's, twitter's live 
> status feed with HBase is possible? (Distribution of feeds to friend lists 
> etc.)
>

I can't really answer that, as I haven't used it in this way. It has a
really good write throughput though. You could optimise for this case.
In memory tables etcetera. You'd want to not run any map reduce over
it at the same time.

> Or
> Its main goal is to analyse whole data and calculations for the internal use? 
> Not  for serving them to users in realtime like RDBMS?

HBase 0.20 can handle real time now. See
http://devblog.streamy.com/2009/07/24/streamy-hadoop-summit-hbase-goes-realtime/


>
> Thank you so much.


hope that helps.
~Tim

Reply via email to