Blobs Fees Adjustments

[gabriele-0201]

This PR contains an implementation of the following formula for fees:

inclusion_fee = base_fee + [targeted_length_fee_adjustment * length_fee] + [targeted_weight_fee_adjustment * weight_fee];

This is accomplished by using LengthToFee dynamically along with FeeMultiplierUpdate.

Decoupling length_fee and weight_fee could be the perfect solution to handle both blockchain usage congestion and blocks usage.

In #16, I explain how difficult it is to find a relationship between those two. The main difference is that each submit_blob extrinsic has a weight even with a blob of zero size.

Having two adjustment factors can handle every scenario. Some extreme ones are:

• Very low general usage, few transactions with small blobs => both targeted_length_fee_adjustment and targeted_weight_fee_adjustment will decrease
• A few transactions, each with very large blobs => only targeted_length_fee_adjustment will increase while targeted_weight_fee_adjustment won't, because the blob size has a minimal impact on the submit_blob weight
• (the one that I couldn't handle with Blobs Fee Adjustment #143) A lot of transactions with very small blobs => in this case, decoupling is the key. The block size will remain small, so targeted_length_fee_adjustment won't increase. However, with many transactions, weight has a huge impact, making targeted_weight_fee_adjustment the key to handling chain congestion.

TODOs

• Refactor, move adjustment fees logic into a separate module
  • possibly PR into substrate to avoid dealing with weird side effects
• Unit testing
• Define AdjustmentVariables (at most 30% per hour in total)
• Integration testing, make sure everything works with lots of transactions and can easily recover from low usage

[gabriele-0201]

Current dependencies on/for this PR:

• main
  • PR Blobs Fees Adjustments #149 👈
    • PR Refactor Blobs Fees Adjustments #150
      • PR Unit Testing Blobs Fees Adjustments #151
        • PR AdjustmentVariable Blobs Fees Adjustments #152
          • PR Fee Adjustment Tests #153
            • PR Dynamic Fee Adjustment Crate #161

This stack of pull requests is managed by Graphite.

[rphmeier]

Needs docs (btw, this is 5* 1024 * 1024 because it's the maximum allowed PoV size).

[gabriele-0201]

Previously, RuntimeBlockLength was defined like this:

pub const RuntimeBlockLength: BlockLength = BlockLength::max_with_normal_ratio(5 * 1024 * 1024, NORMAL_DISPATCH_RATIO);

I decided to make the value a constant because I needed it inside the BlobsLengthToFee logic. I will add docs to describe what the constant represents.

I'm not sure why 5 * 1024 * 1024 was used as the maximum block size. I have an idea, but I'm not entirely sure.

[pepyakin]

What size is used in Substrate as the default maximum block size? I am 80% sure it's 4 MiB.

[gabriele-0201]

Both in substrate node template and cumulus node template is used 5 * 1024 * 1024 as BlockLength

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
docs-site	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Dec 27, 2023 6:18pm