Add tx weighting stage

[sakridge]

Problem

Transaction don't have the stake weight of the sender and to order them it would be nice to have.

Summary of Changes

Add a stage to add the stake weights after streamer and before sigverify.

Fixes #

[buffalu]

do top validators know this is coming? hopefully they’re running more than 1G NICs lol

[buffalu]

also this could make DOS worse without special forwarder software?

not only can you screw with block propagation of current leader by stealing bandwidth from them, but also can forward my DOS packets through the highest staker to the current leader

[sakridge]

also this could make DOS worse without special forwarder software?

not only can you screw with block propagation of current leader by stealing bandwidth from them, but also can forward my DOS packets through the highest staker to the current leader

That high staker which is forwarding will QoS those packets and will prefer to send packets from other staked nodes first before it sends the unstaked DoS packets.

[segfaultdoc]

also this could make DOS worse without special forwarder software?
not only can you screw with block propagation of current leader by stealing bandwidth from them, but also can forward my DOS packets through the highest staker to the current leader

That high staker which is forwarding will QoS those packets and will prefer to send packets from other staked nodes first before it sends the unstaked DoS packets.

Summary of Changes

Why do you assume DoS packets will come from unstaked nodes?
Also, should those changes to the QoS service be added to this PR?

[codecov]

Codecov Report

Merging #21953 (a01fb1a) into master (ba046bd) will decrease coverage by 0.0%.
The diff coverage is 84.1%.

@@            Coverage Diff            @@
##           master   #21953     +/-   ##
=========================================
- Coverage    81.1%    81.1%   -0.1%     
=========================================
  Files         555      556      +1     
  Lines      150664   150742     +78     
=========================================
+ Hits       122300   122351     +51     
- Misses      28364    28391     +27     

[michaelh-laine]

How do tx first get into "the system" as most non-vote tx are first submitted to/via RPC nodes which have zero stake. Does this PR not create a scenario where votes push out tx from all other clients? And if there is a sequence of low-staked leaders that receive a burst of client transactions (as RPC forwards to upcoming leaders) they then receive low priority when forwarded (if not immediately included in a block).

Perhaps there's more intelligence in the mechanism that would address these issues, would appreciate some clarity on this.

(Question derived from discussion on Validator roundtable call on 20.01.2022)

[tao-stones]

How do tx first get into "the system" as most non-vote tx are first submitted to/via RPC nodes which have zero stake. Does this PR not create a scenario where votes push out tx from all other clients?

In short, no, votes will not push out low-staked non-vote transactions.
Votes and non-vote transactions are in separate leader threads, with separate queues. The stake weighted shuffle applies to each queue independently.

[tao-stones]

Close this PR as it has been merged into #22958, reviews and comments should be made on combined PR.