BUIP proposal: Integrate ATMP bottleneck solutions from Gigaperf to release client

[reina]

HI! I haven't BUIPed before, I just wanted to suggest this, because it would be great for BU I think, if one day the improvements that were developed during the Giga Block test initiative that are on the gigaperf branch, will be applied to the release client too. I hope @theZerg and everyone could give some insight on this : ) @solex .

Below are the details, I used the format @torusJKL has used:

BUIP: Implement ATMP bottleneck solutions from Giga test client to release client
Date: 2.9.2018
Proposer: reina

This BUIP proposes to test and inplement the ATMP bottleneck solutions from Gigablock test initiative (currently existing on gigaperf branch of BU) onto the BU release clients.

Motivation
Apply the findings and code work from the GigaBlock Test Initiative to the main BU release stream, to improve the handling of sustained tx flow in BU software and on Bitcoin Cash nework, improving efficiency and capabilities. This will invite more users to choose BU software for their node and be an improvement for existing BU users too.

Development task
Integrate the ATMP bottleneck solutions from Gigablock test initiative onto main BU branch and test.

 

[adamstgbit]

what solution?
they wrote some code to help with ATMP?
I would be interested to see what this solution is.

 

[solex]

Hi @reina,

As it happens, we can proceed to add some specific giga_perf branch (GTI) improvements to the BU dev version without a BUIP as they come under pre-approved categories:

1) In the case of Accept to Memory Pool bottleneck issues, fixing them is a scaling improvement, which goes all the way back to the BU articles and BUIP001. So, we are fine to proceed as soon as development priorities allow.

2) General maintenance which is a continuous process happening in the background all the time.

Development BUIPs are for functional changes which alter the usage of the client in a manner which which some users will like, but others may not.

I can confirm that @theZerg is now looking at getting some of the GTI improvements into dev, though I assume it is not a simple merge as we can imagine that changes done to facilitate the record 1GB test blocks were quick and fast, but now need to be re-implemented to a higher development standard. Because BCH txn volumes have been typically quite small for a long time, priorities have been elsewhere.

 

[reina]

Hi @solex and @theZerg , that's awesome then! Thanks for the update!