For a parachain to be added to Polkadot it must inhabit one of the available parachain slots. A parachain slot is a scarce resource on Polkadot and only a limited number will be available. As parachains ramp up, there may only be a few slots that are unlocked every few months. The goal is to eventually have 100 parachain slots available on Kusama (these will be split between parachains and the parathread pool). If a parachain wants to have guaranteed block inclusion at every Relay Chain block, it must acquire a parachain slot.
The parachain slots will be sold according to an unpermissioned candle auction that has been slightly modified to be secure on a blockchain.
Mechanics of a Candle auction
Candle auctions are a variant of open auctions where bidders submit bids that are increasingly higher and the highest bidder at the conclusion of the auction is considered the winner.
Candle auctions were originally employed in 16th century for the sale of ships and get their name from the "inch of a candle" that determined the open period of the auction. When the flame extinguished and the candle went out, the auction would suddenly terminate and the standing bid at that point would win.
When candle auctions are used online, they require a random number to decide the moment of termination.
Parachain slot auctions differ slightly from a normal candle auction in that it does not use the random number to decide the duration of its opening phase. Instead, it has a known open phase and will be retroactively determined (at the normal close) to have ended at some point in the past during the ending phase. So during the open phase, bids will continue to be accepted, but later bids have higher probability of losing since the retroactively determined close moment may be found to have preceded the time that a bid was submitted.
Randomness in Action
The following example will showcase the randomness mechanics of the candle auction for the ninth auction on Kusama. Keep in mind that the candle phase has a uniform termination profile and has an equal probability of ending at any given block, and the termination block cannot be predicted before or during the auction.
- Auction 9 starts at
The auction has a full duration equal to
block 72000is the "ending period", which is divided into 3600 samples of 20 blocks. Figuratively, the candle is lit, and the candle phase lasts for 72,000 blocks.
- The winning sample during the ending period turned out to have the
Sample 1078 refers to the winner as of
block 9362014 + 21560, which equals
- The parent block was a new BABE session in the 'Logs', which updated the randomness that was used to
select that sample index.
You'd be able to inspect the state at the end of
block 9434277and see the sample indices with an archive node. The digest in the 'Logs' of
9434277is decodable and contains the random value as well as the BABE authorities.
- As a result, the winner of this auction did not turn out to be the highest bid during the full duration.
The open and transparent nature of blockchain systems opens attack vectors that are non-existent in traditional auction formats. Normal open auctions in particular can be vulnerable to auction sniping when implemented over the internet or on a blockchain.
Auction sniping takes place when the end of an auction is known and bidders are hesitant to bid their true price early, in hopes of paying less than they actually value the item.
For example, Alice may value an item at auction for 30 USD. She submits an initial bid of 10 USD in hopes of acquiring the items at a lower price. Alice's strategy is to place incrementally higher bids until her true value of 30 USD is exceeded. Another bidder Eve values the same item at 11 USD. Eve's strategy is to watch the auction and submit a bid of 11 USD at the last second. Alice will have no time to respond to this bid before the close of the auction and will lose the item. The auction mechanism is sub-optimal because it has not discovered the true price of the item and the item has not gone to the actor who valued it the most.
On blockchains this problem may be even worse, since it potentially gives the producer of the block an opportunity to snipe any auction at the last concluding block by adding it themselves and/or ignoring other bids. There is also the possibility of a malicious bidder or a block producer trying to grief honest bidders by sniping auctions.
For this reason, Vickrey auctions, a variant of second price auction in which bids are hidden and only revealed in a later phase, have emerged as a well-regarded mechanic. For example, it is implemented as the mechanism to auction human readable names on the ENS. The Candle auction is another solution that does not need the two-step commit and reveal schemes (a main component of Vickrey auctions), and for this reason allows smart contracts to participate.
Candle auctions allow everyone to always know the states of the bid, but not when the auction will be determined to have ended. This helps to ensure that bidders are willing to bid their true bids early. Otherwise, they might find themselves in the situation that the auction was determined to have ended before they even bid.
Polkadot will use a random beacon based on the VRF that's used also in other places of the protocol. The VRF will provide the base of the randomness, which will retroactively determine the end-time of the auction.
The slot durations are capped to 1 year and divided into 6-week periods; Parachains may lease a slot for any combination of periods of the slot duration. Parachains may lease more than one slot over time, meaning that they could extend their lease to Polkadot past the maximum duration by leasing a contiguous slot.
Note: Individual parachain slots are fungible. This means that parachains do not need to always inhabit the same slot, but as long as a parachain inhabits any slot it can continue as a parachain.
Parachains, or parachain teams, can bid in the auction by specifying the slot range that they want to lease as well as the number of tokens they are willing to reserve. Bidders can be either ordinary accounts, or use the crowdloan functionality to source tokens from the community.
Parachain slots at genesis
Slot A | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |...
Slot B | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |...
Slot C |__________| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |...
Slot D |__________| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |...
Slot E |__________|___________| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |...
Each period of the range 1 - 4 represents a 6-week duration for a total of 1 year
Bidders will submit a configuration of bids specifying the token amount they are willing to bond and
for which periods. The slot ranges may be any of the periods 1 -
n is the number of
periods available for a slot (
n will be 8 for both Polkadot and Kusama).
Please note: If you bond tokens with a parachain slot, you cannot stake with those tokens. In this way, you pay for the parachain slot by forfeiting the opportunity to earn staking rewards.
A bidder configuration for a single bidder may look like the following pseudocode example:
const bids = [
range: [1, 2, 3, 4, 5, 6, 7, 8],
range: [1, 2, 3, 4],
range: [2, 3, 4, 5, 6, 7],
The important concept to understand from this example is that bidders may submit different
configurations at different prices (
bond_amount). However, only one of these bids would be
eligible to win exclusive of the others.
The winner selection algorithm will pick bids that may be non-overlapping in order to maximize the amount of tokens held over the entire lease duration of the parachain slot. This means that the highest bidder for any given slot lease period might not always win (see the example below).
A random number, which is based on the VRF used by Polkadot, is determined at each block. Additionally, each auction will have a threshold that starts at 0 and increases to 1. The random number produced by the VRF is examined next to the threshold to determine if that block is the end of the auction within the so-called ending period. Additionally, the VRF will pick a block from the last epoch to take the state of bids from (to mitigate some types of attacks from malicious validators).
There is one parachain slot available.
75 for the range 1 - 8.
100 for the range 5 - 8.
40 for the range 1 - 4.
Let's calculate each bidder's valuation according to the algorithm. We do this by multiplying the bond amount by the number of periods in the specified range of the bid.
Charlie - 75 * 8 = 600 for range 1 - 8
Dave - 100 * 4 = 400 for range 5 - 8
Emily - 40 * 4 = 160 for range 1 - 4
Although Dave had the highest bid in accordance to token amount, when we do the calculations we see
that since he only bid for a range of 4, he would need to share the slot with Emily who bid much
less. Together Dave's and Emily's bids only equals a valuation of
Charlie's valuation for the entire range is
600. Therefore Charlie is awarded the complete range of
the parachain slot.
Why doesn't everyone bid for the max length?
For the duration of the slot the tokens bid in the auction will be locked up. This means that there are opportunity costs from the possibility of using those tokens for something else. For parachains that are beneficial to Polkadot, this should align the interests between parachains and the Polkadot Relay Chain.
How does this mechanism help ensure parachain diversity?
The method for dividing the parachain slots into intervals was partly inspired by the desire to allow for a greater amount of parachain diversity, and prevent particularly large and well-funded parachains from hoarding slots. By making each period a 6-week duration but the overall slot a 1-year duration, the mechanism can cope with well-funded parachains that will ensure they secure a slot at the end of their lease, while gradually allowing other parachains to enter the ecosystem to occupy the durations that are not filled. For example, if a large, well-funded parachain has already acquired a slot for range 1 - 8, they would be very interested in getting the next slot that would open for 2 - 9. Under this mechanism that parachain could acquire just the period 9 (since that is the only one it needs) and allow range 2 - 8 of the second parachain slot to be occupied by another.
Why is randomness difficult on blockchains?
Randomness is problematic for blockchain systems. Generating a random number trustlessly on a transparent and open network in which other parties must be able to verify opens the possibility for actors to attempt to alter or manipulate the randomness. There have been a few solutions that have been put forward, including hash-onions like RANDAO and verifiable random functions (VRFs). The latter is what Polkadot uses as a base for its randomness.
Are there other ways of acquiring a slot besides the candle auction?
Another way, besides the candle auction, to acquire a parachain slot is through a secondary market where an actor who has already won a parachain slot can resell the slot along with the associated deposit of tokens that is locked up to another buyer. This would allow the seller to get liquid tokens in exchange for the parachain slot and the buyer to acquire the slot as well as the deposited tokens.
A number of system or common-good parachains may be granted slots by the governing bodies of the Relay Chain. System parachains can be recognized by a parachain ID lower than 1_000, and common-good parachains by a parachain ID between 1_000 and 1_999. Other parachains will have IDs 2_000 or higher. Such parachains would not have to bid for or renew their slots as they would be considered essential to the ecosystem's future.
- Parachain Allocation - W3F research page on parachain allocation that goes more in depth to the mechanism
- Research Update: The Case for Candle Auctions - W3F breakdown and research update about candle auctions
- Front-Running, Smart Contracts, and Candle Auctions W3F Research team discusses how to remedy current blockchain auction setbacks with candle auctions