Original title: Behind the scenes of Solana's slow validator problem
Original author: Jack Kubinec, Blockworks
Original translation: 深潮 TechFlow
As the code issues of Solana have gradually been resolved over the past few years, block times (the time required for the network to generate new transaction blocks) have significantly decreased, even dropping below the nominal 400 milliseconds.
However, an interesting trend has emerged over the past month: the median block time (a key performance metric in blockchain networks that reflects the speed at which the blockchain network processes transactions and generates blocks) has surged, indicating that Solana's speed of adding new transactions to the blockchain is slowing down. The reason lies in a new validator strategy, suggesting that slow block generation may be more profitable. According to Blockworks, Anza, Jito, and Marinade are considering solutions to this issue.
Each Solana block has a validator acting as the leader — responsible for collecting transactions, creating blocks, and broadcasting them to the network. The leader earns transaction fees by creating blocks. More order flow means more fee opportunities, so validators may choose to process transactions in 500 milliseconds rather than 300 milliseconds to increase their earnings.
From a basic level, some Solana validators seem to be waiting as long as possible to package more transactions into blocks to maximize returns. This behavior has led to an increase in the cycle length of Solana.
This is clearly not an ideal situation for a network that aims to be as fast as Nasdaq. Additionally, the reduced cycle means fewer compounding opportunities for staking rewards, a point raised by Max Kaplan, CTO of Sol Strategies.
Solana offers a mechanism called "grace ticks," which is a delay period that allows leaders to successfully submit blocks. This mechanism is designed to prevent unfair penalties for validators located remotely but also opens the door for validators to intentionally delay block submissions.
Additionally, Solana's alternative client Frankendancer recently released a revenue-maximizing scheduler.
According to Kaplan, validators running the client appear to be packaging blocks at a slightly slower than normal rate. However, Kaplan added that compared to more severely delayed validators, the delay of Frankendancer is negligible, and he does not see it as a "bad thing." Moreover, block delay is not a new concept on proof-of-stake blockchains. However, the upgrade of Firedancer may make this strategy more noticeable on Solana. Jump has not commented on this.
Interestingly, Firedancer software engineer Michael McGee described this phenomenon in this week's Lightspeed podcast. He mentioned, "One thing we see with current validators is that they often create more profitable blocks by delaying transaction execution."
Blockworks Research analyst Victor Pham pointed out that those Solana validators that are more noticeably delaying blocks are typically running modified versions of the Agave-Jito client.
For example, in the 802nd cycle in mid-June, the median block time of both Galaxy and Kiln exceeded 570 milliseconds. According to Solana Compass data, some unmarked validators also run slower, while the median block time for Temporal's validators is 475 milliseconds.
Kiln co-founder Ernest Oppetit acknowledged that its validator — the sixth largest staking validator on the Solana network — had delayed block slots for a time but stated that this behavior has now ceased.
"At Kiln, we pride ourselves on providing the highest staking annual percentage yield (APY) in the market without sacrificing safety. We have been researching different parts of the tech stack, including time strategies, and maintaining ongoing discussions with clients, client teams, and foundations. Currently, we follow the protocol and do not delay blocks anymore, but many other validators still do. We believe that ultimately, the incentive issues (where fast block generation leads to reduced rewards) need to be addressed at the protocol level," said Oppetit.
Ben Coverston, Director of Engineering at Temporal, said in response to questions about their validators apparently participating in the slow block trend: "I can say that we are not making people aware of the reasons for this phenomenon."
A spokesperson for Galaxy stated, "As a service provider, we support validator configurations that can prioritize maximizing client staking rewards. On Solana, this may mean proposing slightly slower blocks to ensure higher reward acquisition. Galaxy has also been responsive to community feedback and has adjusted block times to an acceptable range."
However, the Solana validator community generally believes that slowing down the network is inappropriate, and slow validators are currently facing strong public opposition.
They may soon face more substantial penalties. According to Blockworks, Jito plans to blacklist slow validators from its staking pool, which is the largest on the Solana network.
Brian Smith, chairman of the Jito Foundation, stated that the organization is "drafting a governance proposal to empower a committee to remove underperformers from the JitoSOL delegation pool. This proposal should be open for community discussion in a few days."
Michael Repetny, co-founder of Marinade, stated that the staking pool provider is "considering submitting this issue as a governance proposal to discuss the pros and cons of classifying [slow validators] as a hard rule/violation of delegation strategy."
Protocol-level solutions are also being advanced. Anza's GitHub repository shows a new proposal to reduce Solana's grace tick period by half. Additionally, the consensus mechanism reform proposed by Solana is expected to address this issue.
"Alpenglow will address this issue by enabling a skip voting feature," said Brennan Watt, Vice President of Engineering at Anza.
Watt revealed in a recent episode of the Lightspeed podcast that Anza hopes to launch Alpenglow on the mainnet before the Solana Breakpoint conference in December this year.
Original link
