using Riak TS for KV buckets

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

using Riak TS for KV buckets

Gal Ben Haim
is using KV bucket operations in a Riak TS cluster the same as in a Riak KV cluster 
(in terms of performance) ?

I had a phone call with a solution architect from Basho who said it fits only lightweight KV operations. This is written nowhere in the docs..

The reason I'm asking about that is the Spark connector, that works only with Riak TS. The solution architect also said that KV queries from the Spark connector are experimental, which is also written nowhere in the docs..

I'd really like to get some more insights about this

_______________________________________________
riak-users mailing list
[hidden email]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
Reply | Threaded
Open this post in threaded view
|

Re: using Riak TS for KV buckets

Pavel Hardak
Hi Gal,
I am the product manager for Riak TS and Spark Connector. Thanks for your questions and PRs.
We have two versions of Riak - KV and TS. They share most of the code, but RIak TS is optimised for time series operations, while Riak KV - for key value operations. It is correct that most KV operations are available in TS, but there are some caveats (TS supports only LevelDB backend, KV also supports bitcask and memory) and several KV features are disabled (for example AAE and commit hooks). The untended usage of KV features in Riak TS is to support time-series and IoT use cases - e.g. sensors metadata or device registry. 
In terms of performance - it very much depends on the specific use case and what features you plan to use. Generally we tune Riak TS to better support TS workloads, so if you want to squeze maximal perforance in KV operations, then Riak KV is the better choice. For developmemt and testing purposes Riak TS can be used as KV "stand in". It might work really well, but KV functionality inside TS was not tested as throughfully as we do it for KV. As I said above, several features are disabled, so some tests are failing. This is why it is not "officially" supported for deployment.
I will ask to add this clarificaltion to the documentation.  
Spark Connector currently supports Riak TS - both TS tables and KV buckets. Spark Connector is not officially supported yet with Riak KV . Some operations will work, some will not. There is new fuctionality (for KV buckets), which was added to Riak TS, and will be added to Riak KV as well. This is planned for the next verison of Riak KV, 2.3. 
Hope it asnwers your question.
Best regards,
Pavel Hardak

Date: Thu, 1 Dec 2016 08:54:46 +0200
From: Gal Ben Haim <[hidden email]>
To: [hidden email]
Subject: using Riak TS for KV buckets
is using KV bucket operations in a Riak TS cluster the same as in a Riak KV cluster (in terms of performance) ?
I had a phone call with a solution architect from Basho who said it fits
only lightweight KV operations. This is written nowhere in the docs..

The reason I'm asking about that is the Spark connector, that works only
with Riak TS. The solution architect also said that KV queries from the
Spark connector are experimental, which is also written nowhere in the
docs..

I'd really like to get some more insights about this








_______________________________________________
riak-users mailing list
[hidden email]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com