add module fee_adjustment

sugondat-chain/runtimes/sugondat-kusama/src/fee_adjustment.rs

@@ -0,0 +1,193 @@
+use crate::constants::kusama::currency::MILLICENTS;
+use frame_support::parameter_types;
+use pallet_transaction_payment::{Multiplier, MultiplierUpdate, TargetedFeeAdjustment};
+use sp_runtime::{
+    traits::{Bounded, Convert},
+    FixedPointNumber, Perquintill, SaturatedConversion, Saturating,
+};
+use sp_weights::Weight;
+use sugondat_primitives::{Balance, MAXIMUM_BLOCK_LENGTH};
+
+parameter_types! {
+    /// Relay Chain `TransactionByteFee` / 10
+    pub const TransactionByteFee: Balance = MILLICENTS;
+
+    // parameters used by BlobsFeeAdjustment
+    // to update NextFeeMultiplier and NextLengthMultiplier
+    //
+    // Common constants used in all runtimes for SlowAdjustingFeeUpdate
+    /// The portion of the `NORMAL_DISPATCH_RATIO` that we adjust the fees with. Blocks filled less
+    /// than this will decrease the weight and more will increase.
+    pub storage TargetBlockFullness: Perquintill = Perquintill::from_percent(25);
+
+    /// The adjustment variable of the runtime. Higher values will cause `TargetBlockFullness` to
+    /// change the fees more rapidly.
+    pub AdjustmentVariableBlockFullness: Multiplier = Multiplier::saturating_from_rational(75, 1_000_000);
+    /// that combined with `AdjustmentVariable`, we can recover from the minimum.
+    /// See `multiplier_can_grow_from_zero`.
+    pub MinimumMultiplierBlockFullness: Multiplier = Multiplier::saturating_from_rational(1, 10u128);
+    /// The maximum amount of the multiplier.
+    pub MaximumMultiplierBlockFullness: Multiplier = Bounded::max_value();
+
+
+    pub storage NextLengthMultiplier: Multiplier = Multiplier::saturating_from_integer(1);
+    pub storage TargetBlockSize: Perquintill = Perquintill::from_percent(16); // 0.8MiB
+    // TODO: update those value accordingly with https://github.com/thrumdev/blobs/issues/16
+    pub AdjustmentVariableBlockSize: Multiplier = Multiplier::saturating_from_rational(75, 1_000_000);
+    pub MinimumMultiplierBlockSize: Multiplier = Multiplier::saturating_from_rational(1, 10u128);
+    pub MaximumMultiplierBlockSize: Multiplier = Bounded::max_value();
+}
+
+/// Currently pallet_transaction_payment use the following formula:
+///
+/// ```ignore
+/// inclusion_fee = base_fee + length_fee + [targeted_fee_adjustment * weight_fee];
+/// ```
+///
+/// Letting us able to update `targeted_fee_adjustment` at the end of each block
+/// thanks to `FeeMultiplierUpdate`, this associated type is called inside the `on_finalize`
+/// of the transaction_payment pallet with the aim of converting the before `targeted_fee_adjustment`
+/// to a new one based on the congestion of the network
+///
+/// What this struct does is this PLUS a side effect, the goal is to reach a different formula to
+/// calculate fees:
+///
+/// ```ignore
+/// inclusion_fee = base_fee + [targeted_length_fee_adjustment * length_fee] + [targeted_weight_fee_adjustment * weight_fee];
+/// ```
+///
+/// As you can see `targeted_fee_adjustment` becomes `targeted_weight_fee_adjustment` but the behavior
+/// remains the same, the side effect is the changing to the value `targeted_length_fee_adjustment`,
+/// this formula is achievable because inside pallet_transaction_payment the function `compute_fee_raw`
+/// that just computes the final fee associated with an extrinsic uses the associated type `LengthToFee`
+/// that converts the length of an extrinsic to a fee.
+///
+/// By default the implementation is a constant multiplication but we want to achieve a dynamic formula
+/// that can adapt based on the usage of the network, this can't solely be done by this struct but needs
+/// to be bundled with a custom implementation of `LengthToFee`.
+///
+/// This struct ONLY provide a dynamic update of `targeted_length_fee_adjustment` and `targeted_weight_fee_adjustment`
+/// based on the congestion and usage of the blocks, while the formula si effectively implemented like
+/// explained above only thanks to `LengthToFee`
+pub struct BlobsFeeAdjustment<T: frame_system::Config>(core::marker::PhantomData<T>);
+
+impl<T: frame_system::Config> Convert<Multiplier, Multiplier> for BlobsFeeAdjustment<T>
+where
+    T: frame_system::Config,
+{
+    /// This function should be a pure function used to update NextFeeMultiplier
+    /// but will also has the side effect of update NextLengthMultiplier
+    fn convert(previous_fee_multiplier: Multiplier) -> Multiplier {
+        // Update NextLengthMultiplier
+
+        // To update the value will be used the same formula as TargetedFeeAdjustment,
+        // described here: https://research.web3.foundation/Polkadot/overview/token-economics#2-slow-adjusting-mechanism
+        //
+        // so this is mainly a copy paste of that function because it works on normalized mesurments,
+        // so if it is ref_time, proof_size or length of the extrinsic the mutliplier will be evaluated properly.
+        // The main problem is that TargetedFeeAdjustment::convert uses directly a call to the storage to extract
+        // the weight of the current block so there is no way to pass the length as input argument,
+        // here I will copy paste all the needed part to update properly NextLengthMultiplier
+
+        // Defensive only. The multiplier in storage should always be at most positive. Nonetheless
+        // we recover here in case of errors, because any value below this would be stale and can
+        // never change.
+
+        let previous_len_multiplier = NextLengthMultiplier::get();
+        let min_multiplier = MinimumMultiplierBlockSize::get();
+        let max_multiplier = MaximumMultiplierBlockSize::get();
+        let previous_len_multiplier = previous_len_multiplier.max(min_multiplier);
+
+        // Pick the limiting dimension. (from TargetedFeeAdjustment::convert)
+        //
+        // In this case it is the length of all extrinsic, always
+        let (normal_limiting_dimension, max_limiting_dimension) = (
+            <frame_system::Pallet<T>>::all_extrinsics_len(),
+            MAXIMUM_BLOCK_LENGTH as u64,
+        );
+
+        let target_block_size = TargetBlockSize::get();
+        let adjustment_variable = AdjustmentVariableBlockSize::get();
+
+        let target_size = (target_block_size * max_limiting_dimension) as u128;
+        let block_size = normal_limiting_dimension as u128;
+
+        // determines if the first_term is positive
+        let positive = block_size >= target_size;
+        let diff_abs = block_size.max(target_size) - block_size.min(target_size);
+
+        // defensive only, a test case assures that the maximum weight diff can fit in Multiplier
+        // without any saturation.
+        let diff = Multiplier::saturating_from_rational(diff_abs, max_limiting_dimension.max(1));
+        let diff_squared = diff.saturating_mul(diff);
+
+        let v_squared_2 = adjustment_variable.saturating_mul(adjustment_variable)
+            / Multiplier::saturating_from_integer(2);
+
+        let first_term = adjustment_variable.saturating_mul(diff);
+        let second_term = v_squared_2.saturating_mul(diff_squared);
+
+        let new_len_multiplier = if positive {
+            let excess = first_term
+                .saturating_add(second_term)
+                .saturating_mul(previous_len_multiplier);
+            previous_len_multiplier
+                .saturating_add(excess)
+                .clamp(min_multiplier, max_multiplier)
+        } else {
+            // Defensive-only: first_term > second_term. Safe subtraction.
+            let negative = first_term
+                .saturating_sub(second_term)
+                .saturating_mul(previous_len_multiplier);
+            previous_len_multiplier
+                .saturating_sub(negative)
+                .clamp(min_multiplier, max_multiplier)
+        };
+
+        NextLengthMultiplier::set(&new_len_multiplier);
+
+        // Update NextFeeMultiplier
+        //
+        // Here is the tricky part, this method return the new value associated with
+        // NextFeeMultiplier (in the old fashion) because weight dynamic adjustment is battle tested
+        // while previously have updated the `NextLengthMultiplier` used in `LengthToWeight`
+        TargetedFeeAdjustment::<
+            T,
+            TargetBlockFullness,
+            AdjustmentVariableBlockFullness,
+            MinimumMultiplierBlockFullness,
+            MaximumMultiplierBlockFullness,
+        >::convert(previous_fee_multiplier)
+    }
+}
+
+impl<T: frame_system::Config> MultiplierUpdate for BlobsFeeAdjustment<T> {
+    fn min() -> Multiplier {
+        MinimumMultiplierBlockFullness::get()
+    }
+    fn max() -> Multiplier {
+        MaximumMultiplierBlockFullness::get()
+    }
+    fn target() -> Perquintill {
+        TargetBlockFullness::get()
+    }
+    fn variability() -> Multiplier {
+        AdjustmentVariableBlockFullness::get()
+    }
+}
+
+pub struct BlobsLengthToFee<T: frame_system::Config>(core::marker::PhantomData<T>);
+
+impl<T: frame_system::Config> sp_weights::WeightToFee for BlobsLengthToFee<T> {
+    type Balance = Balance;
+
+    fn weight_to_fee(weight: &Weight) -> Self::Balance {
+        // really weird but weight.ref_time will contain the length of the extrinsic
+        let length_fee = Self::Balance::saturated_from(weight.ref_time())
+            .saturating_mul(TransactionByteFee::get());
+        let multiplier = NextLengthMultiplier::get();
+
+        // final adjusted length fee
+        multiplier.saturating_mul_int(length_fee)
+    }
+}