HBql query performance

[Joan Han <jo...@gmail.com>]

Hi,

Has anyone used HBql ? I don't see many discussion on this in the mailing
list. Thought to ask around to see if anyone has opinion on the usage.

Here is my question:

1) Can HBql be used as real time analytic query ?

I have a HBase table (10 millions rows)  with many columns and only 1 column
family. Can I do something like this ?

select * from table where cf.age > 30 AND cf.title='david horwitz' AND price
> 6.99 ? ( condition can be more here. Note this is not the HBql syntax)

Will this query give near real time response ?

2) If not what other HBase client choices do we have to achieve near real
time query performance using multiples clauses with AND, OR, NOT etc ?


Thanks

-Joan

Re: HBql query performance

[Jean-Daniel Cryans <jd...@apache.org>]

Inline.

J-D

On Mon, Sep 26, 2011 at 4:14 PM, Joan Han <jo...@gmail.com> wrote:
> Hi,
>
> Has anyone used HBql ? I don't see many discussion on this in the mailing
> list. Thought to ask around to see if anyone has opinion on the usage.

Yeah not a lot.

>
> Here is my question:
>
> 1) Can HBql be used as real time analytic query ?

No.

>
> I have a HBase table (10 millions rows)  with many columns and only 1 column
> family. Can I do something like this ?
>
> select * from table where cf.age > 30 AND cf.title='david horwitz' AND price
>> 6.99 ? ( condition can be more here. Note this is not the HBql syntax)

No, Hive could do something like that but in a batch fashion.

>
> Will this query give near real time response ?

No.

>
> 2) If not what other HBase client choices do we have to achieve near real
> time query performance using multiples clauses with AND, OR, NOT etc ?

It's not a matter of client, it's the architecture. There are no
indexes in HBase (except for the row key).  No relations. The only
fast access patterns are random reads (of a single row) and sequential
reads of a few thousand rows.

>
>
> Thanks
>
> -Joan
>