bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 2쪽 중 2쪽
levant issue, it’s important to note that is possible to completely clear the remaining
balance by either 1. execute another transfer, which will most likely transfer the remaining dust balance as it
would be too small to accrue interest in a reasonably short amount of time, or 2. redeem the dust balance and
transfer the underlying asset.
2. Interest stream can only be redirected if there is a principal balance: This means that only accounts
that have a principal balance Bp can redirect their interest. If users redeem or transfer everything, their
interest redirection is reset. As a side effect of this, interest generated only by the redirected balance Br cannot
be redirected.
16
4 Stable Rate Theory
The following chapter explains how the stable rates are applied to the system and the limitations.
Implementation of a fixed rate model on top of a pool is complicated. Indeed, fixed rates are hard to handle
algorithmically, as the cost of borrowing money varies with market conditions and the liquidity available. There
might therefore be situations (sudden market changes, bank runs ...) in which handling stable rate borrow positions
would need using specific heuristics based on time or economical constraints. Following this reasoning, we identified
two possible ways of handling fixed rates:
1. Imposing time constraints : fixed rates might work perfectly fine in a time constrained fashion. If a loan
has a stable duration, it should survive extreme market conditions, as the borrower must repay at the end of
the loan period. Unfortunately, time constrained fixed rate loans aren’t suitable for our specific use case of
open ended loan. It would require a certain degree of UX friction where users would need to create and handle
multiple loans with different times constraints.
2. Imposing rates constraints : An interest rate calculated at the beginning of a loan might be impacted by
market conditions, keeping it from staying fixed. If the rate diverges too much from the market, it can be
readjusted. This would not be a pure fixed rate, open term loan - as the rate might vary throughout the loan
duration – yet users will experience actual fixed rates during specific time periods, or when there is enough
liquidity available. This particular implementation has been chosen to be integrated into Aave’s Protocol
under the name stable rate .
4.1 Lending Rate Oracle
Figure 13: Lending Rate Oracle
The first component to be integrated into the Protocol protocol is a Lending Rate Oracle, which will provide
information to the contracts on the actual market rates that other lending platforms, both centralized and
decentralized, are providing. The average market lending rate Mr is defined for i platforms with P ir the lending rate
and P iv the borrowing volume:
Mr =
∑n
i=1 P ir P iv∑n
i=1 P iv
The market rate will be updated daily, initially by Aave.
17
4.2 Current Stable Borrow Rate Rs
The current stable borrow rate is calculated as follows:
Rts =
{
Mr + UUoptimal Rslope1, if U≤ Uoptimal
Mr + Rslope1 + U −Uoptimal1−Uoptimal Rslope2, if U > U optimal
With:
- Mr the average market lending rate.
- Rslope1 the interest rate slope below Uoptimal, increases the rate as U increases.
- Rslope2 the interest rate slope beyond Uoptimal, increases as the difference between U and Uoptimal increases.
- U is the utilization rate.
Note: Rs does NOT impact existing stable rates positions – this is applied only to new opened positions.
4.3 Limitations on Stable Rate Positions
To avoid abuses on stable rate loans, the following limitations have been applied to the stable rate borrowing model:
1. Users cannot deposit as collateral more liquidity than what they are trying to borrow. Eg. a user deposits 10
million DAI collateral, tries to borrow 1 million DAI. This is to prevent the following attack vector:
Given: Bs = 18%AP R, Mr = 9%AP R, Rl = 12%AP R
Users might try to artificially lower Bs to the value of Mr by depositing a huge amount of liquidity which
would cause Bs to drop, then borrow from the same liquidity at a lower rate, withdraw the liquidity previously
deposited to cause Bs and the liquidity rate Rl to raise again; then finally deposit the amount borrowed to
earn interest on the previously borrowed funds. Although this attack can still be carried out using multiple
accounts, this particular constraint makes the attack more complicated as it requires more money (and a
different collateral currency). This works well in combination with the interest rate rebalancing in the next
section.
2. Borrowers will only be able to borrow up to Tr of the available liquidity at the current borrow rate. So, for
every specific value of Bs, there is only up to Tr of liquidity available for a single borrower. This is to avoid
that a specific borrower would borrow too much available liquidity at a too competitive rate.
4.4 Stable Rate Rebalancing
The last and perhaps most important constraint of the stable rate model is the rate rebalancing. This is to work
around changes in market conditions or increased cost of money within the pool.
The stable rate rebalancing will happen in two specific situations:
1. Rebalancing up . The stable rate of a user x is rebalanced to the most recent value of Bs when a user could
earn interest by borrowing:
Bxs < R l with Bxs the stable borrow rate of user x
2. Rebalancing down . The stable rate of a user x is rebalanced to the most recent value of Bs, if:
Bxs > B s(1 + ∆B s)
with ∆ B s a rate delta established by governance which defines the window above Bs to rebalance interest
rates. If a user pays too much interest beyond that range, the rate is balanced down.
18
4.5 The Rebalancing Process
The LendingPool contract exposes a function rebalanceStableBorrowRate(address reserve, address user)
which allows to rebalance the stable rate interest of a specific user. Anybody can call this function: however, there
isn’t any direct incentive for the caller to rebalance the rate of a specific user. For this reason, Aave will provide
an agent that will periodically monitor all the stable rates positions and rebalance the ones that will be deemed
necessary. The rebalance strategy will be decided offchain by the agent, this means that users that satisfy the
rebalance conditions may not be rebalanced immediately. Since those conditions depend on the liquidity avail-
able and the state of market, there might be some transitory situations in which an immediate rebalance is not needed.
This does not add any element of centralization to the protocol. Even if the agent stops working, anybody can call
the rebalance function of the LendingPool contract. Although there isn’t any direct incentive in doing it (“why
should I do it?”) there is an indirect incentive for the ecosystem. In fact, even if the agent should cease to exist,
depositors might still want to trigger a rebalance up of the lowest borrow rate positions, to increase the liquidity
rate and/or force borrowers to close up their positions, increasing the available liquidity. In case of a rescale down,
instead, borrowers have a direct incentive in performing a rebalance of their positions to lower the interest rate.
The following flowchart explains the sequence of actions of the function rebalanceStableBorrowRate(). The
compounded balance that is accumulated until the instant at which the rebalance happens, is not affected by the
rebalance.
19
Figure 14: Rebalancing
20
5 Conclusion
The Aave Protocol relies on a lending pool model to offer high liquidity. Loans are backed by collateral and
represented by aTokens, derivative tokens which accrue the interests. The parameters such as interest rate and
Loan-To-Value are token specific.
Aave improves Decentralized Finance’s current offering, bringing two key innovations to the lending ecosystem:
• Stable Rates to help borrowers’ financial planning;
• Flash Loans to borrow without collateral during a single transaction.
Following the launch of the mainnet, Aave will uphold its commitment to decentralization through additional features.
The Pool Factory will allow anyone to launch their own lending pool based on our smart-contracts. Governance will
be on-chain with rights represented by:
• The LEND token at Protocol level for updates of the smart contract;
• aTokens at Pool level for pool specific parameters.
21프로젝트로 돌아가기