Hesitation between three Elasticsearch-based Architectures

[bechir-jamoussi]

Hello everyone,
Let me present to you my Use Case:
I am trying to develop an algorithm that chooses the best communication channel(SMS, RCS, Messenger, Whatsapp, E-mail) and the best moment to communicate with the clients. There are two phases for this algorithm:

• The prediction phase in which it will, based on usage and segmentation criteria, pick the right channel.

• The adaptation phase, in which it will receive everyday logs (CDRs) of the client's actions and usages and try to adapt.(Machine Learning Process)

I am hesitating between three Elasticsearch-based architectures, can you please give me some advice, and if you have any documentation or Architecture to suggest please feel free.

1st: Logstash(Producer) -> Kafka -> Logstash(Consumer) -> Elasticsearch -> Eland(Machine Learning)

2nd: Logstash -> Spark -> Elasticsearch -> Eland(Machine Learning)

3rd: Logstash -> Kafka -> Spark (Do Machine Learning) -> Elasticsearch

Thank you in advance.

[sholzhauer]

I'm not familiar with Spark and Eland. However I am pretty sure for anyone to help you they will need to understand your (planned) usage of the elastic stack.

What usecase(s) do you plan to carry out?

[bechir-jamoussi]

Thank you for your comment, I have just edited my post.