Freqtrade cluster

[bigtcze]

Hi, today in a new freqtrade version were introduced freqtrade clusters. Will you consider implementing this into your strategy? https://www.freqtrade.io/en/latest/producer-consumer/

[iterativv]

Hi. Sure thing, let me look into it first.

[DarkByteZero]

I would really like this implemented, running like 4 FreqTrade Instances with NFI is crazy intensive on the CPU.

[Toxiiin]

Did anybody already tried to use the consumer/producer logic? Thought about splitting the workload into a pipeline like:

flowchart TD
    A[Pairlist Producer] -->|Fetch global pairlist| B(OffsetFilter 0:20)
    A[Pairlist Producer] -->|Fetch global pairlist| C(OffsetFilter 21:40)
    A[Pairlist Producer] -->|Fetch global pairlist| D(OffsetFilter 41:60)
    A[Pairlist Producer] -->|Fetch global pairlist| E(OffsetFilter 61:80)

    B --> |Pairlist subset| F[Indicator Producer 0:20]
    C --> |Pairlist subset| G[Indicator Producer 21:40]
    D --> |Pairlist subset| H[Indicator Producer 41:60]
    E --> |Pairlist subset| I[Indicator Producer 61:80]
    
    F --> |Fetch indicators| J[Indicator Consumer]
    G --> |Fetch indicators| J[Indicator Consumer]
    H --> |Fetch indicators| J[Indicator Consumer]
    I --> |Fetch indicators| J[Indicator Consumer]

Any plans on integrating this (optionally)? Found some attempts on Discord, but nothing "official".

Had a small break, started again with my pi 4b to check current situation and recognized, that X4 is pretty much more computationally intensive than the version I used in '22 ... 😆 Everybody already switched to x86-64 or Pi 5 for whitelists with pairs >40?