What happened to Riak-TS and are there any plans?

[Ridderby]

Hi,

I am evaluating practices for data storage in my Erlang echo system and Riak thus OpenRiak seems really nice. I am both seeking to store Key-Value data and Time Series data.

The documentation of Riak-TS really appeals to me but I cannot find TS among the OpenRiak repos so I wounder what has been decided and if there are any plans for Riak TS to be part of OpenRiak?

Or are there perhaps another strategy for storing Time Series Data using OpenRiak instead?

Thanks in advance,
Erik

[martinsumner]

There has been no maintenance of the Riak TS branches through the OpenRiak community, and updating Riak TS is not presently part of the community roadmap. There are non-trivial problems to resolve to maintain Riak TS going forward, in particular the dependency on the leveldb fork.

As for a strategy for storing time-series data within Riak KV itself. There have been real-world examples of Riak KV being used at a significant scale for some specific time-series-related requirements, with Riak storing batches of time series data. There has been other interest in seeing this support formalised, and some suggestions put forward as to how the Query API could be expanded to make this more viable. However, there's no immediate push to prioritise this at the moment, but I'm sure there would be interest within the community in hearing more about potential use cases.

[Ridderby]

Thanks @martinsumner ,

The use case at hand is about home automation and storing measures from solar power systems with or without batteries for future analysis. The long term case is to use AI to predict when to charge batteries or sell existing capacity of batteries to grid given weather fore case and grid prices (that can be negative). Here in Sweden we will also have efficiency rates to consider which complicate things further. The intention is to have everything on site and there are no plans to provide as service to others, but it is open source for others to run their own stuff at location.

The software is BatMan (Battery Manager) and it can be found at gitlab. The erlang backend is not usable right now though, the C++ backend works pretty well here and we use it daily.

Anyhow, the core scenario is to collect measures and at small or even tiny scale compared to big data alternatives. Yet hobby (or perhaps research) projects like this is where you explore technology for later use in professional software.

[martinsumner]

Riak is a scale-out system, so normally we don't see it used unless in clusters of multiple (5+) nodes, where there is requirement for resilience of data to device failure. I'm not sure if this fits with your local storage model. The nodes can be small devices though, it works well with a redundant set of inexpensive compute nodes.

Good luck with your plans.

[Ridderby]

Thanks.

The business case I have beside home projects (where I learn stuff) may actual need to scale out a bit. Don't know how much yet, depends on how well we manage to get customer :)

[hmmr]

Post Basho, in 2022, Riak TS had been uplifted to then-current versions of otp and rebased onto current develop branches in riak_kv, riak_core by TI Tokyo; packages of riak_ts-3.0.0 (with features that were slated for 1.6.0 in Basho sequence, to be released in spring of 2017) are available in https://files.tiot.jp/riak/ts/3.0/3.0.0/.