If my application supports multiple websites (multi-tenancy), what is the
best way of naming my keys for each column family?

example, say I have a column family "articles".

Now these articles could be for 100's of websites, how would I go about
naming my keys for this table?

Retrieving a single article by key is fairly straight forward, but what if I
want to get the last 10 articles, it would make sense to have a key that
grouped these together somehow correct?

It sounds like you would probably want a composite key that includes a
unique identifier for each site plus a timestamp (possibly reversed)
which is discussed in Tom White's book and Jonathan Gray's blog post:
http://devblog.streamy.com/2009/04/23/hbase-row-key-design-for-paging-limit-offset-queries/

Cheers,
Sarah

On Tue, Jul 13, 2010 at 11:17 PM, S Ahmed <[EMAIL PROTECTED]> wrote:
> If my application supports multiple websites (multi-tenancy), what is the
> best way of naming my keys for each column family?
>
> example, say I have a column family "articles".
>
> Now these articles could be for 100's of websites, how would I go about
> naming my keys for this table?
>
> Retrieving a single article by key is fairly straight forward, but what if I
> want to get the last 10 articles, it would make sense to have a key that
> grouped these together somehow correct?
>

--
Sarah Sproehnle
Educational Services
Cloudera, Inc
http://www.cloudera.com/training

Great that clears things up, so its about converting things to a bytes.

On Tue, Jul 13, 2010 at 11:22 PM, Sarah Sproehnle <[EMAIL PROTECTED]>wrote:

> It sounds like you would probably want a composite key that includes a
> unique identifier for each site plus a timestamp (possibly reversed)
> which is discussed in Tom White's book and Jonathan Gray's blog post:
>
> http://devblog.streamy.com/2009/04/23/hbase-row-key-design-for-paging-limit-offset-queries/
>
> Cheers,
> Sarah
>
> On Tue, Jul 13, 2010 at 11:17 PM, S Ahmed <[EMAIL PROTECTED]> wrote:
> > If my application supports multiple websites (multi-tenancy), what is the
> > best way of naming my keys for each column family?
> >
> > example, say I have a column family "articles".
> >
> > Now these articles could be for 100's of websites, how would I go about
> > naming my keys for this table?
> >
> > Retrieving a single article by key is fairly straight forward, but what
> if I
> > want to get the last 10 articles, it would make sense to have a key that
> > grouped these together somehow correct?
> >
>
>
>
> --
> Sarah Sproehnle
> Educational Services
> Cloudera, Inc
> http://www.cloudera.com/training
>

Can' seem to find a page in the wiki that goes over the various data types a
'key' can be?  Or does it always have to be a byte array?

On Tue, Jul 13, 2010 at 11:47 PM, S Ahmed <[EMAIL PROTECTED]> wrote:

> Great that clears things up, so its about converting things to a bytes.
>
>
> On Tue, Jul 13, 2010 at 11:22 PM, Sarah Sproehnle <[EMAIL PROTECTED]>wrote:
>
>> It sounds like you would probably want a composite key that includes a
>> unique identifier for each site plus a timestamp (possibly reversed)
>> which is discussed in Tom White's book and Jonathan Gray's blog post:
>>
>> http://devblog.streamy.com/2009/04/23/hbase-row-key-design-for-paging-limit-offset-queries/
>>
>> Cheers,
>> Sarah
>>
>> On Tue, Jul 13, 2010 at 11:17 PM, S Ahmed <[EMAIL PROTECTED]> wrote:
>> > If my application supports multiple websites (multi-tenancy), what is
>> the
>> > best way of naming my keys for each column family?
>> >
>> > example, say I have a column family "articles".
>> >
>> > Now these articles could be for 100's of websites, how would I go about
>> > naming my keys for this table?
>> >
>> > Retrieving a single article by key is fairly straight forward, but what
>> if I
>> > want to get the last 10 articles, it would make sense to have a key that
>> > grouped these together somehow correct?
>> >
>>
>>
>>
>> --
>> Sarah Sproehnle
>> Educational Services
>> Cloudera, Inc
>> http://www.cloudera.com/training
>>
>
>

HBase only speaks byte arrays, so your keys can be anything (binaries,
strings, integers, anything).

J-D

On Tue, Jul 13, 2010 at 9:34 PM, S Ahmed <[EMAIL PROTECTED]> wrote:
> Can' seem to find a page in the wiki that goes over the various data types a
> 'key' can be?  Or does it always have to be a byte array?
>
> On Tue, Jul 13, 2010 at 11:47 PM, S Ahmed <[EMAIL PROTECTED]> wrote:
>
>> Great that clears things up, so its about converting things to a bytes.
>>
>>
>> On Tue, Jul 13, 2010 at 11:22 PM, Sarah Sproehnle <[EMAIL PROTECTED]>wrote:
>>
>>> It sounds like you would probably want a composite key that includes a
>>> unique identifier for each site plus a timestamp (possibly reversed)
>>> which is discussed in Tom White's book and Jonathan Gray's blog post:
>>>
>>> http://devblog.streamy.com/2009/04/23/hbase-row-key-design-for-paging-limit-offset-queries/
>>>
>>> Cheers,
>>> Sarah
>>>
>>> On Tue, Jul 13, 2010 at 11:17 PM, S Ahmed <[EMAIL PROTECTED]> wrote:
>>> > If my application supports multiple websites (multi-tenancy), what is
>>> the
>>> > best way of naming my keys for each column family?
>>> >
>>> > example, say I have a column family "articles".
>>> >
>>> > Now these articles could be for 100's of websites, how would I go about
>>> > naming my keys for this table?
>>> >
>>> > Retrieving a single article by key is fairly straight forward, but what
>>> if I
>>> > want to get the last 10 articles, it would make sense to have a key that
>>> > grouped these together somehow correct?
>>> >
>>>
>>>
>>>
>>> --
>>> Sarah Sproehnle
>>> Educational Services
>>> Cloudera, Inc
>>> http://www.cloudera.com/training
>>>
>>
>>
>