4 ms·

Why I Am Excited about Riak Search

48 points by pharkmillups 16y ago

siculars 16y ago
I gave two talks[1] on Riak in NYC this month and one of the main points tht I tried to stress was that Riak is more than a monolithic persistent data store. Riak is actually a very well thought out federation of independant code modules. This had not always been as evident when looking at earlier releases of the source. Over time what became clear is tht Basho was refactoring their code to allow for interesting combinations and enhancemnts over their standard offering. As the OP points out, the recent modularization of riak_core has unleashed a wave of potential within the Riak ecosystem as recently manifested in the form of Riak Search. I'm am certain this design approach of code modularization will prove immensely valuable as Riak continues to mature. [1]http://www.slideshare.net/siculars/adding-riak-to-your-nosql-bag-of-tricks http://www.slideshare.net/siculars/adding-riak-to-your-nosql...
ericflo 16y ago
Riak Search is a great example of the result of getting the fundamentals right up-front, proving and improving them in a concrete product (Riak), and then extracting the best bits into a reusable library (Riak Core). The way they've done this matches my own philosophy on how this stuff should be done. Not to mention that Riak Search is very cool in and of itself.
roder 16y ago
Riak Search is the first public demonstration that Riak Core is a meaningful base on which to build distributed systems beyond just a key/value store. By using the same central code base for distribution, dispatch, ownership, failure management, and node administration, we are able to confidently make many of the same guarantees for Search that we have made all along for Riak's key/value storage. Even if Search itself wasn't such a compelling product, it is exciting as a proof of the value of Riak Core Riak Core can abstract away the difficulties of decentralized computing is really exciting to me as well. It's a decentralized message queue or apache zookeeper written in Erlang. I really think the basho team has done an amazing job modularizing the components Riak is comprised of (core, kv, search, etc) to make all bits really re-usable. Well done Basho!
jeffreymcmanus 16y ago
It's great that they're adding this (full-text search is one of the big missing features from most NoSQL products), but I wonder why it wouldn't make more sense to just provide great integration into one of the existing dedicated search services like Sphinx?
- nl 16y ago
  No, because they want it to be one of those dedicated search products like Sphinx. At the moment, there are two main open source products people look at when they implement full text search: Sphinx and Lucene/Solr. I'm not a Sphinx expert, but I know Lucene/Solr quite well. One feature Solr is currently working on is called Solr Cloud (http://wiki.apache.org/solr/SolrCloud http://wiki.apache.org/solr/SolrCloud), which is a Zookeeper co-ordinated clustered Solr implementation, which should make it a lot easier to automatically resize your Solr cluster depending on usage. (At the moment you can do this, but it requires rsync scripts etc etc, and is kind of messy). It looks to me like Riak Search is aiming for this market, which I think is a good strategy. (Note that Solr is often used as a NoSQL database, so in that way Riak & Solr are direct competitors)

48 points by pharkmillups 16y ago

I gave two talks[1] on Riak in NYC this month and one of the main points tht I tried to stress was that Riak is more than a monolithic persistent data store. Riak is actually a very well thought out federation of independant code modules. This had not always been as evident when looking at earlier releases of the source. Over time what became clear is tht Basho was refactoring their code to allow for interesting combinations and enhancemnts over their standard offering. As the OP points out, the recent modularization of riak_core has unleashed a wave of potential within the Riak ecosystem as recently manifested in the form of Riak Search. I'm am certain this design approach of code modularization will prove immensely valuable as Riak continues to mature. [1]http://www.slideshare.net/siculars/adding-riak-to-your-nosql-bag-of-tricks http://www.slideshare.net/siculars/adding-riak-to-your-nosql...

ericflo 16y ago

Riak Search is a great example of the result of getting the fundamentals right up-front, proving and improving them in a concrete product (Riak), and then extracting the best bits into a reusable library (Riak Core). The way they've done this matches my own philosophy on how this stuff should be done. Not to mention that Riak Search is very cool in and of itself.

roder 16y ago

Riak Search is the first public demonstration that Riak Core is a meaningful base on which to build distributed systems beyond just a key/value store. By using the same central code base for distribution, dispatch, ownership, failure management, and node administration, we are able to confidently make many of the same guarantees for Search that we have made all along for Riak's key/value storage. Even if Search itself wasn't such a compelling product, it is exciting as a proof of the value of Riak Core Riak Core can abstract away the difficulties of decentralized computing is really exciting to me as well. It's a decentralized message queue or apache zookeeper written in Erlang. I really think the basho team has done an amazing job modularizing the components Riak is comprised of (core, kv, search, etc) to make all bits really re-usable. Well done Basho!

jeffreymcmanus 16y ago

It's great that they're adding this (full-text search is one of the big missing features from most NoSQL products), but I wonder why it wouldn't make more sense to just provide great integration into one of the existing dedicated search services like Sphinx?

nl 16y ago
No, because they want it to be one of those dedicated search products like Sphinx. At the moment, there are two main open source products people look at when they implement full text search: Sphinx and Lucene/Solr. I'm not a Sphinx expert, but I know Lucene/Solr quite well. One feature Solr is currently working on is called Solr Cloud (http://wiki.apache.org/solr/SolrCloud http://wiki.apache.org/solr/SolrCloud), which is a Zookeeper co-ordinated clustered Solr implementation, which should make it a lot easier to automatically resize your Solr cluster depending on usage. (At the moment you can do this, but it requires rsync scripts etc etc, and is kind of messy). It looks to me like Riak Search is aiming for this market, which I think is a good strategy. (Note that Solr is often used as a NoSQL database, so in that way Riak & Solr are direct competitors)