You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@hbase.apache.org by Peter Haidinyak <ph...@local.com> on 2010/12/04 01:28:10 UTC

Which approach would be better

Hi,
  Which would be a better approach.

1.       Having  every entry into HBase use a unique Row Key

2.       Having similar entries into HBase use the same Row Key and then use versions to extract the data.

I have noticed that option 2 is much slower for putting data into HBase by a factor of 2.5 but would extracting the information be faster?

Thanks

-Pete

Re: Which approach would be better

Posted by 明珠刘 <re...@gmail.com>.

2. I'm afraid not.


2010/12/4 Peter Haidinyak <ph...@local.com>

> Hi,
>  Which would be a better approach.
>
> 1.       Having  every entry into HBase use a unique Row Key
>
> 2.       Having similar entries into HBase use the same Row Key and then
> use versions to extract the data.
>
> I have noticed that option 2 is much slower for putting data into HBase by
> a factor of 2.5 but would extracting the information be faster?
>
> Thanks
>
> -Pete
>

Re: Which approach would be better

Posted by Stack <st...@duboce.net>.

What do you mean by similar?

I'd think the speed would be the same doing inserts.  How many rows
and regions when you are done?  What size cluster?

How do you intend to query HBase?  Will you be requesting clumps of
'similars' or just getting an item at a time?

St.Ack

On Fri, Dec 3, 2010 at 4:28 PM, Peter Haidinyak <ph...@local.com> wrote:
> Hi,
>  Which would be a better approach.
>
> 1.       Having  every entry into HBase use a unique Row Key
>
> 2.       Having similar entries into HBase use the same Row Key and then use versions to extract the data.
>
> I have noticed that option 2 is much slower for putting data into HBase by a factor of 2.5 but would extracting the information be faster?
>
> Thanks
>
> -Pete
>