You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@storm.apache.org by Aaron Bossert <ma...@gmail.com> on 2016/03/09 22:38:59 UTC
Re: "Prediction" with known values
This seems like a problem that can be solved any number of ways without a database. Since storm bolts can be written in just about any language, pick your favorite and work it that way.
Your description of your data is a bit confusing though...perhaps if you post a sample of the data, it might make it easier for someone to give you a suggestion.
Also, the problem you are trying to solve is not really a storm issue, but a programming logic issue, you might also have better luck asking that part of the question on stack overflow or a similar site...just a thought.
--
Aaron
> On Mar 9, 2016, at 4:31 PM, Maria Musterfrau <da...@gmx.at> wrote:
>
> Hello
>
> I have a question regarding Storm. I would like to predict the next hour in realtime. But it is not really a prediction as I already know the values. I receive a message which tells me that a program has been started. Every program has a list of values (every minute a different value for different time frames depending on the program), stored in an external database. I would like to add up all the values of one minute over all started programs but I do not know how to solve the issue with the timestamp. Because I only have a timestamp for the start message, but not for the individual values. So how do I know which values belong to which minute. Here is an example to show it:
>
> p1 p2 p3 p4 sum/minute
> minute 1 3 7 = 10
> minute 2 4 5 6 = 15
> minute 3 5 7 6 8 = 26
> minute 4 7 8 4 1 = 20
> minute 5 8 5 5 = 18
> ...
>
> Is this possible within a Storm bolt? Or would it only be possible with a NoSQL database or any other tools?
> Does anyone have any ideas?
>
> Thank you in advance.
>
> Regards,
> Daniela