譯文作者：0xEvan, Primev

原文標(biāo)題：Blob 市場(chǎng)中的審查、延遲和預(yù)先確認(rèn)

來(lái)源：鏡子

編譯：Franci，ETHconomics 研究空間

譯者導(dǎo)讀

前面我們發(fā)的幾篇文章都和 blob 交易本身以及 4844 費(fèi)用相關(guān)，而在這篇文章機(jī)制中，作者通過(guò)對(duì)過(guò)去一年的數(shù)據(jù)回測(cè)，以模擬 blob 市場(chǎng)的潛力——它可以承載多少數(shù)據(jù)？能否滿足rollup的數(shù)據(jù)可用需求？

除了blob空間的供需曲線對(duì)blob市場(chǎng)會(huì)造成影響之外，驗(yàn)證者的時(shí)間博弈以及構(gòu)建者的審查也對(duì)這個(gè)市場(chǎng)帶來(lái)一些負(fù)面影響。本文對(duì)可能產(chǎn)生的blob交易廣播延遲進(jìn)行有了數(shù)據(jù)分析，由此產(chǎn)生的用戶體驗(yàn)、匯總數(shù)據(jù)可用開(kāi)銷影響進(jìn)行了轉(zhuǎn)型，也提出了可能的解決方案——預(yù)確認(rèn) blob 交易。

總的來(lái)說(shuō)，這篇文章是比較全面的blob市場(chǎng)分析，讀者大可以通過(guò)文章的脈絡(luò)，探索未來(lái)blob空間的格局。

正文：

特別感謝@terencechain的審閱、@BertKellerman的意見(jiàn)和@ethpandaops提供的holesky測(cè)試網(wǎng)數(shù)據(jù)。

長(zhǎng)話短說(shuō)

• 我們的研究深入探討了新興的EIP-4844 blob市場(chǎng)，該市場(chǎng)的運(yùn)行方式與EIP-1559的gas定價(jià)方式類似。不同的是，它直接針對(duì)區(qū)塊鏈建設(shè)者的小費(fèi)機(jī)制，因而沒(méi)有激勵(lì)其儲(chǔ)備blob 交易，這可能導(dǎo)致 blob 交易體驗(yàn)不穩(wěn)定，以及預(yù)留進(jìn)區(qū)塊有一定的挑戰(zhàn)。

• 我們注意到，雖然 blob 交易容量很大（約 125 kB）且比同等大小的通話數(shù)據(jù)便宜，但它們大大增加了以太坊區(qū)塊的大小。這也意味著 blob 交易給區(qū)塊帶來(lái)了增量的競(jìng)價(jià)能力。

• 我們利用這個(gè)新市場(chǎng)的容量能夠吸收當(dāng)前rollup的數(shù)據(jù)需求，將標(biāo)準(zhǔn)區(qū)塊空間的gas成本降低15-20%，從而解鎖了更明顯的mev機(jī)會(huì)。

• 我們觀察到，在網(wǎng)絡(luò)活動(dòng)增加的時(shí)期，blob 交易可能會(huì)使區(qū)塊廣播速度減慢幾千幾千的數(shù)量級(jí)，這可能導(dǎo)致區(qū)塊構(gòu)建者為了在 mev-boost 中維持具有競(jìng)爭(zhēng)性的競(jìng)價(jià)并審查 blob 。

• 我們?cè)u(píng)估，“預(yù)確認(rèn)出價(jià)”可以緩解這些挑戰(zhàn)，blob預(yù)確認(rèn)可以增強(qiáng)EIP-4844的能力，為L(zhǎng)2用戶提供更好的交易體驗(yàn)，以及為rollups提供穩(wěn)定的交易備用體驗(yàn)。

• 我們將在Holesky測(cè)試網(wǎng)上進(jìn)行，收集區(qū)塊構(gòu)建者數(shù)據(jù)，并通過(guò)mev-commit將relay作為blob預(yù)確認(rèn)成功，我們邀請(qǐng)PBS實(shí)驗(yàn)的相關(guān)參與者來(lái)一起進(jìn)行實(shí)驗(yàn)。

引入

EIP-4844引入了一個(gè)blob市場(chǎng)，從而擴(kuò)展了以太坊的數(shù)據(jù)可用性。這個(gè)新興市場(chǎng)使用類似于EIP-1559的gas價(jià)格機(jī)制來(lái)定價(jià)和匯率blob的基本gas費(fèi)用。然而，與type2交易不同的是，blob 市場(chǎng)中用戶沒(méi)有直接的方式出價(jià)給構(gòu)建者作為，節(jié)約其 blob 的小額費(fèi)用。缺乏優(yōu)先收費(fèi)的設(shè)計(jì)使得準(zhǔn)確定價(jià) blob 節(jié)約費(fèi)用變得困難。此外，預(yù)計(jì) blob 的區(qū)塊在網(wǎng)絡(luò)中的廣播中會(huì)變慢，因?yàn)?blob 是以太坊交易類型中最大的一種。如果構(gòu)建者在一個(gè)區(qū)塊中接受許多 blob，他們可能會(huì)面臨更高的區(qū)塊重組風(fēng)險(xiǎn)，因此，假設(shè)構(gòu)建者是經(jīng)濟(jì)學(xué)方面理性的人，其可能會(huì)選擇在mev高峰期審查blob，以保持塊構(gòu)建的低延遲。

我們提出了一個(gè)與 blob 相關(guān)的區(qū)塊構(gòu)建和 mev-boost 數(shù)據(jù)收集工作，以及一個(gè)使用 mev-commit 的 blob 預(yù)確認(rèn)實(shí)驗(yàn)，并邀請(qǐng)社區(qū)中的 rollup、relay、區(qū)塊構(gòu)建者和發(fā)起者參與我們對(duì) EIP-4844 中 blob 相關(guān)行為的意見(jiàn)表明，L1 blob 預(yù)確認(rèn)可以增強(qiáng) blob 市場(chǎng)帶來(lái)的功效，為 L2 用戶提供更好的交易體驗(yàn)，為 rollup 在出現(xiàn) mev 條件下提供可靠的資源體驗(yàn)，以及給rollup為中心的以太坊路線圖提供一個(gè)更穩(wěn)定的未來(lái)。

理解 Blob 市場(chǎng)

斑點(diǎn)交易

EIP-4844 引入了一種 type3 交易（稱為 blob 交易）。進(jìn)行 blob 的交易亦類似于常規(guī)交易，但增加了 blob 數(shù)據(jù)、KZG 承諾和證明。與標(biāo)準(zhǔn)以太坊交易相比，Blob 極大(約125 kB)，且比等量的calldata便宜。Calldata的價(jià)格為每個(gè)非零字節(jié)可16gas且大小變化；而blob數(shù)據(jù)的價(jià)格為每字節(jié)1.04gas，且每個(gè)blob上限為131,072氣體。

斑點(diǎn)氣體機(jī)制

Blob 基礎(chǔ)氣體定價(jià)隨著定價(jià)網(wǎng)絡(luò)的堵塞情況的費(fèi)用而變化，這與 EIP-1559 類似。主要機(jī)制區(qū)別在于，blob 基礎(chǔ)氣體的價(jià)格基于 Blob 使用數(shù)量，而 EIP-1559 是基于上個(gè)區(qū)塊氣體的利用率變化（使用的gas數(shù)量之于目標(biāo)gas數(shù)量而言）。目標(biāo)Blob的數(shù)量是3（0.375 MB），每個(gè)區(qū)塊的最大數(shù)量是6（0.75 MB）。Blob基礎(chǔ)gas價(jià)格簡(jiǎn)單設(shè)置為1 wei 。

當(dāng)提交一個(gè)blob交易時(shí)，發(fā)送者將提交max_fee_per_blob_gas作為他們?cè)敢鉃閎lob基礎(chǔ)gas費(fèi)支付的最高價(jià)格，所有這些費(fèi)用都將被納入考慮。max_fee_per_blob_gas與type0和type2交易中的max_fee_per_gas類似。如果用戶想提交額外的費(fèi)用以預(yù)算其交易為主，那么他們還會(huì)提交max_priority_fee。然而，max_priority_fee僅覆蓋交易中非blobgas部分。也就是說(shuō)，這種blobgas費(fèi)用機(jī)制下用戶不能直接提交blob預(yù)算小費(fèi)給構(gòu)建者。

（原文者注：關(guān)于4844費(fèi)用機(jī)制原理分析，我們社區(qū)撰寫了更詳細(xì)的文章，請(qǐng)看這里）

Blob 市場(chǎng)容量

在本節(jié)中，我們對(duì)2023年1月到2024年1月期間rollup網(wǎng)絡(luò)的歷史交易活動(dòng)進(jìn)行了回測(cè)，以展示blob市場(chǎng)的結(jié)構(gòu)。我們關(guān)注以太坊上最活躍的rollup的交易，并利用歷史數(shù)據(jù)它來(lái)了一個(gè)模擬實(shí)時(shí)的 blob 市場(chǎng)。當(dāng)然，這個(gè)市場(chǎng)正在積極增長(zhǎng)，且 blob 尚未在主網(wǎng)上線，論文使用了整個(gè) 2023 年的歷史數(shù)據(jù)，旨在模擬的潛力。

Based on the historical calldata activity of rollup and simulating its use in the block space of type3 transactions, we can see that the blob market price can easily absorb all the capacity of the rollup without causing the blob market price to exceed the blob base gas minimum value ( That is 1 wei).

Figure: base blob gas per block

Although the rollup publishes more data to Ethereum, most blocks are still below the target blob number, which ensures that blob gas prices remain low.

Figure: The lighter the color, the more times a block that packs a specific number of blobs is built.

This means that the calldata overhead of the blob market will be lower (calldata consumes 16 gas per byte, while blob consumes 1 gas per byte), and the gas price will also be lower (the gas price of calldata is gwei level, while blob The gas price is at the wei level), thus saving two additional layers of costs for rollup.

Not only can the blob market easily absorb the data availability needs of current rollups, it can also enable non-blob markets to free up more block space and reduce gas overhead by more than 15-20%. The reduction in gas overhead in turn increases the bidding power of users/searchers, builders, and validators, and unlocks new mev opportunities that were priced out prior to EIP-4844.

Figure: Impact of EIP-4844 on standard block space (based on 2023 data)

Rollup requires more data availability

Rollup greatly affects the gas usage in the block. They are currently the largest type of gas users in the Ethereum block space. In 2023, rollup stored a record amount of transaction data on Ethereum, as shown in the figure below:

Figure: Calldata saved on Ethereum hits record high

The daily average chart below shows that rollups are starting to occupy more than 15% of each block they are in, directly affecting the price of use for other users.

This may worsen in the event of a black swan and increased demand. As recently as December 2023, the inscription craze caused excessive transaction volume, causing Arbitrum’s sequencer to be offline for about an hour. When the Arbitrum sequencer resumed operations and began publishing the backlog of saved state data, the sequencer immediately monopolized the entire block space, causing the gas price to soar to more than 140 gwei, consuming up to 90% of the gas in all blocks, making The network was unavailable to most users for several hours.

In the next section, we discuss how timing games and censorship might impact this market even in the absence of this surge in demand.

Challenges Facing the Blob Market: A Review

Broadcasting of Blob transactions

EIP-4844 increases the bandwidth requirements per beacon block by up to ~0.75 MB, 42m gas, to accommodate up to 6 additional blobs per beacon block. Unlike calldata, which is stored permanently, blobs persist in beacon nodes for a short period of time (18 days as of February 2024) to keep the growth of the network's archived state under control.

Additionally, blob transactions have two network representations – a blob transaction for block builders and a blob sidecar for validators. Blob sidecars are designed for forward compatibility.

The blob must first be broadcast through the execution layer and then to the consensus layer. **This means that the builder (not the validator) has the final say on blob packaging. **Proposers can only exclude blob transactions under mev-boost dynamics based on invalid promises/proofs.

Figure: Execution verification is performed by builders, consensus verification is performed by validators

A Block Builder’s Perspective

Recent research on validator “timing games” highlights how latency optimization can strategically enable node operators to maximize profits by delaying block proposals. The authors explain that this is detrimental to the health of the chain. Blob transactions further complicate this game by adding varying amounts of latency when the blob sidecar broadcasts.

Blob transactions are equivalent to the largest possible transaction size type. For this reason, blocks carrying these transactions propagate more slowly, making block builders less competitive in winning mev-boost bids. As a result, this incentivizes block builders to review blobs temporarily or even indefinitely so that they can submit mev bids with greater frequency.

The ethpanda team has been testing real-world latency on testnet using Xatu. They have observers set up in multiple locations around the world (NYC, FRA, BLR, SYD) using different Ethereum consensus clients (Prysm, Nimbus, Lodestar, and Lighthouse) to measure real-world latency. The February 20, 2024 Holesky Blob data snapshot shows that a large amount of latency is generated throughout the mev pipeline.

After a block builder wins a mev-boost bid auction, the proposer must wait for the blob sidecar to be broadcast before validating the blob packaged in the block. The following table shows that with a sample size of 800 blob sidecars, the minimum time for a single blob sidecar broadcast is approximately 400 milliseconds

Chart 1. Blob broadcast time vs. number of blobs contained in a single slot

Figure: The small size of the data is one reason for some of the counterintuitive observations described in this dataset.

The table below shows the difference in latency when waiting for more blob sidecars to arrive. The 50th percentile (p50) in the table shows that the difference in latency between a block carrying 2 blobs and a block carrying 6 blobs is approximately 225 milliseconds.

Chart 2. Time difference between the first arriving and last arriving blob sidecar out of the total number of blob sidecars grouped by block

This blob broadcast delay creates additional block reorganization risk for block builders, who receive minimal financial gain when filling their own blocks with blobs. Builders may choose to exclude/censor blob transactions to avoid potential reorganizations. If a block contains a large number of mevs, economically rational builders need to appropriately compensate for this risk by rolling up the network.

About the user experience of package bidding in the Blob market

In this paper on the study of the validator time game, it is pointed out that later in the mev-boost bidding process, larger bids are associated with larger blocks. As the bid and gas price rise, a larger share of ETH will be destroyed in subsequent slots. If the base fee increases, but the mev withdrawal amount remains the same, the builder has less room to bid on the proposer's future revenue.

Translator's note:

(Value inflow: MEV + handling fee

Value outflow: destroyed ETH + Proposer’s PriorityFee + builder’s tip to Proposer + builder’s own income)

In other words, if more is destroyed, less will be given to the proposer.

Anticipated blob market where capacity exceeds current demand. The destroyed blob base fee will remain on a very small order of magnitude, i.e. tens or hundreds of wei. This is critical for rollups to understand: even if they pay enough base fees, their blob transactions may not be included. The low base fee of the blob market means that blob transactions require bids many times higher to incentivize builders to package such transactions. In this case, the blob transaction will have to be resubmitted at a higher fee, resulting in a poor user experience.

此外，由于EIP-4844下的初始blob市場(chǎng)沒(méi)有打包小費(fèi)機(jī)制（例如blob優(yōu)先gas費(fèi)），這進(jìn)一步加劇了用戶體驗(yàn)問(wèn)題，因?yàn)閞ollups無(wú)法直接競(jìng)價(jià)競(jìng)爭(zhēng)打包blob交易中的空間。

我們看一個(gè)交易示例，假設(shè) Blob 基礎(chǔ) Gas 費(fèi)用為 10 wei，并計(jì)算相同數(shù)據(jù)量的 Blob 開(kāi)銷。

應(yīng)該注意的是，該示例假設(shè)存在可以競(jìng)價(jià) blob 空間的有效包競(jìng)價(jià)機(jī)制。

請(qǐng)看交易樣本：

Calldata - 129,998 字節(jié)（129429 個(gè)非零字節(jié)）~ 10.56 gwei 使用的 2,094,140 Gas（10.55 gwei 基本價(jià)格 + .01 gwei 優(yōu)先費(fèi)）= 0.022 ETH

Blob-128,000bytes~131,072gasusedat1gwei(10weibaseprice+.99999999gweipriorityfee)=0.000131072ETH

計(jì)算得出的結(jié)論是，如果 Rollups 使用 Blob 市場(chǎng)，由于 Blob 基本費(fèi)用較低，他們可能會(huì)提交 100 倍以上的出價(jià)，同時(shí)還可以節(jié)省超過(guò) 150 倍的成本。

較低的 blob baseFee 將允許 rollup 提供更具競(jìng)爭(zhēng)力的出價(jià)，同時(shí)節(jié)省開(kāi)銷。

打包費(fèi)需要與該區(qū)塊現(xiàn)有的機(jī)電項(xiàng)目機(jī)會(huì)一樣具有競(jìng)爭(zhēng)力，以補(bǔ)償建筑商潛在的重組風(fēng)險(xiǎn)，因此即使出價(jià)高出 100 倍也可能不夠。

這是沒(méi)有 blob 預(yù)驗(yàn)證的情況。

通過(guò) mev-commit 進(jìn)行 Blob 預(yù)驗(yàn)證

在這個(gè)時(shí)間博弈中，blob預(yù)確認(rèn)的主要作用是讓一些預(yù)確認(rèn)的blob在mev管道上可用。

通過(guò) mev-commit，每個(gè)預(yù)確認(rèn)提供商對(duì)交易做出自己的承諾。

然后，提供者可以將此數(shù)據(jù)委托給其他參與者（例如塊構(gòu)建器、中繼器、定序器）。

MEV 管道中其他參與者可用的預(yù)確認(rèn)數(shù)據(jù)允許塊構(gòu)建器并行發(fā)送匹配的執(zhí)行負(fù)載。

可以利用這個(gè)概念來(lái)創(chuàng)建預(yù)先確認(rèn)的 blob 裝箱列表，或通過(guò)中繼協(xié)作構(gòu)建 type3 塊空間。

因?yàn)轭A(yù)先確認(rèn)的 blob 是預(yù)先知道的，所以區(qū)塊構(gòu)建者可以在其槽位開(kāi)始之前構(gòu)建攜帶 blob 的未來(lái)區(qū)塊。

這種做法提供了定價(jià)依據(jù)，為建設(shè)強(qiáng)大的期貨市場(chǎng)奠定了基礎(chǔ)。

市場(chǎng)將為rollups提供更可靠的交易打包體驗(yàn)，讓區(qū)塊空間價(jià)格更加穩(wěn)定。

此外，mev-commit 的預(yù)確認(rèn)競(jìng)價(jià)為 rollup 提供了更可靠的價(jià)格發(fā)現(xiàn)機(jī)制，因?yàn)?rollup 可以實(shí)時(shí)更新其預(yù)確認(rèn)競(jìng)價(jià)，而無(wú)需重新提交整個(gè) blob 交易。

最后，捆綁 blob 并使用預(yù)確認(rèn)投標(biāo)允許匯總形成聯(lián)盟。

預(yù)確認(rèn)競(jìng)價(jià)可應(yīng)用于捆綁的 Blob 交易或聚合的 Blob，允許在 Rollups 之間共享競(jìng)價(jià)權(quán)和打包空間，有助于促進(jìn)以太坊 Blob 市場(chǎng)的穩(wěn)定和持續(xù)發(fā)展。

綜上所述

Overall, our research shows that the economics of rollup are improving, while the emergence of new markets requires more factors to be considered, including the impact of time games and the lack of a tipping mechanism. It’s too early to move into the solution phase for the issues we highlighted, but since mev-commit is activated on the Holesky testnet, we can easily conduct experiments with PBS behavioral entities. Primev will collect data on the impact of blobs on block construction and proposer latency, and hopes to understand underlying behavioral patterns.

While economics and user experience are the main drivers for pre-confirmed Type 2 transactions, under EIP-4844, rollups and the rollup-centric transaction packaging experience, reliability and stability of the Ethereum ecosystem will become the key drivers for pre-confirmed blobs. important reason. We will also experiment with a blob pre-confirmation relay, which can leverage blob pre-confirmation and block builder coordination to improve blob sidecar broadcast latency issues on the Holesky testnet. We invite the community to participate in this experiment as it will provide potential solutions for the entire community.

熱點(diǎn)：以太 以太坊 太坊 升級(jí)