Is using the same search index on all CRDT buckets a good idea?

Weixi Yen weixi at spacechatapp.com
Tue Nov 24 01:52:50 EST 2015


Thanks, this is exactly what I was looking for.

On Mon, Nov 23, 2015 at 12:58 PM, Dmitri Zagidulin <dzagidulin at basho.com>
wrote:

> Short answer to 'should I use the same search index across all CRDT
> buckets': Probably not.
>
> Long answer: It depends of what you're going to be storing in your CRDT
> buckets. And what you want to query on.
>
> If your buckets store objects that have fields in common (for example, all
> your CRDTs have a field 'user_id_register'), and you want to query on those
> few fields, then sure, create a custom schema for those fields, and use the
> same index across the buckets.
>
> I'm going to guess that this is not what you intend, though. You're
> probably thinking of either using the default schema (btw - DON'T use the
> default schema in production. It not only indexes -every field- in a
> document, but also -stores the value- for those fields, which is a lot
> slower and wastes a lot more space.), or are going to be using a custom
> schema with some dynamic fields. Something like - index all fields that end
> in *_register, that sort of thing.
>
> Again, that's a bad idea. In relational database terms, this is like
> creating an index on all columns of all tables. It gets expensive very fast.
>
> So, in short:
>
> 1) Don't use the default schema for your indexes in production. It's only
> meant as a developer convenience, while you're learning Solr.
>
> 2) Use custom schemas instead, and only index those fields that you'll be
> querying. So, unless all your CRDTs have common fields, this means creating
> separate custom indexes for your buckets.
>
>
> On Fri, Nov 20, 2015 at 10:36 PM, Weixi Yen <weixi at spacechatapp.com>
> wrote:
>
>> Hello,
>>
>> I have a question about Riak 2.0+ Search best practices.
>>
>> A) I'm wondering if using the same search index on all my CRDT bucket
>> types and then filtering by bucket name is a good idea.
>>
>> B) Or should I be creating different indexes for each bucket if my
>> primary goal is to search by bucket.
>>
>> Are there potentially huge performance issues when going with Option A?
>>
>> Thanks!
>>
>> _______________________________________________
>> riak-users mailing list
>> riak-users at lists.basho.com
>> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.basho.com/pipermail/riak-users_lists.basho.com/attachments/20151123/22b72974/attachment-0002.html>


More information about the riak-users mailing list