Mentioned in This Article
1. Ether.fi makes mump2p mandatory for its operators
On Tuesday, October 20, Optimum, a networking startup led by MIT professor Muriel Médard, announced that its mump2p block-propagation software is running in production on Ethereum mainnet. Liquid restaking provider ether.fi now requires all 12 of its node operators to use it. Optimum says 16 operators in total run the protocol, together securing more than 1.4 million staked ETH.
The decision matters because ether.fi had treated the software as optional until now, according to Optimum. Turning an optional tool into a requirement across an entire validator set is a larger commitment than a pilot, and it places a young piece of networking infrastructure inside one of the biggest staking operations on Ethereum.
Mump2p is not a new blockchain or a new consensus rule. It is a delivery layer for blocks, built to get new data in front of validators sooner than the network's standard peer-to-peer messaging does. Most of the performance claims attached to the announcement come from Optimum itself, and the figures are discussed in detail below.
2. How large is the Optimum network today?
Optimum counts 16 node operators on the protocol, running roughly 22,000 validator keys with more than 1.4 million ETH at stake. The company says that amounts to over 3% of all ETH currently staked. Operators it has named include Blockdaemon, Everstake, P2P.org, Stakefish and Bitcoin Suisse, a mix of infrastructure specialists and a regulated Swiss financial group.
The two operator counts in the announcement are consistent with each other. Twelve is the number of operators working for ether.fi, which are now required to run the software. Sixteen is the wider group on Optimum's network, which includes operators that serve other clients. The mandate therefore covers a portion of the network rather than all of it.
Optimum also cites a figure of about $3.7 billion in what it calls "Total Value Boosted." The company has not defined the metric in the announcement, and the number reflects Optimum's own accounting rather than an independent measurement. Readers should treat it as a marketing statistic until the methodology is explained.
For context, a validator key represents a single staking position of 32 ETH under the standard setup, and large operators run thousands of them from shared infrastructure. Counting keys rather than operators therefore gives a better sense of how much of the network's validation work a software change can touch. Roughly 22,000 keys is a meaningful slice, although it is far from the whole validator set.
3. What is ether.fi and why does its decision carry weight?
Ether.fi is a liquid restaking provider. Users deposit ETH, the protocol delegates it to node operators who run validators, and depositors receive a liquid token that represents their stake. Because the staked ETH is managed through a set of selected operators, the protocol has real influence over what software those operators run.
The company announced a blockspace deal with ETHGas in April. Optimum says ether.fi had previously left mump2p optional for its operators, so the new mandate marks a change in posture rather than a continuation of existing policy.
Rok Kopp, ether.fi's co-founder and chief growth officer, explained the company's approach to infrastructure changes. He said ether.fi does not mandate new infrastructure across its entire validator set unless it is confident the change raises the bar for stakers and for the network. The statement frames the requirement as a judgment that the software improves outcomes, not as a routine upgrade.
The arrangement also illustrates a recurring tension in staking. Delegators chase yield and convenience, while the protocols they use take on the job of choosing infrastructure on their behalf. When a protocol standardizes on one piece of software across all its operators, it gains consistency and a shared performance baseline, but it also ties its operators to the reliability and security of that single vendor's code.
4. Why does block arrival time matter to validators?
Ethereum validators lose attestation and MEV rewards when they receive a block late. Attestations are the votes validators cast on the chain's current state, and MEV, or maximal extractable value, is the additional revenue available from how transactions are ordered inside blocks. A validator that sees a block after its peers does less well on both counts.
Optimum argues that the cost falls most heavily on validators located far from major hosting hubs. Operators in well-connected data centers already see blocks quickly, while those with weaker connectivity trail behind. A mechanism that narrows the gap would, in principle, help the second group the most.
Médard drew a comparison with traditional markets. In her words, firms in traditional finance spend billions shaving milliseconds off execution because speed is money. The analogy presents block latency as a competitive variable for staking operators, much as execution speed is for trading firms.
The effect on any single validator is usually small in a given slot, but it compounds. Ethereum produces a block roughly every twelve seconds, and a validator that is consistently a little behind will, over thousands of slots, accumulate lower returns than a peer that is consistently on time. For large operators managing tens of thousands of keys, even small per-slot differences can add up to meaningful sums over a year.
5. How does mump2p work alongside Ethereum's gossip layer?
According to a blog post Optimum published in August, mump2p runs as an additional channel next to Ethereum's standard gossip layer instead of replacing it. A validator using the software still participates in the usual network and receives blocks through it, but it can also receive the same data through the second path.
The software encodes data using random linear network coding, a technique the company says was co-invented at MIT. In broad terms, network coding lets nodes transmit combinations of data pieces, so a receiver can reconstruct a full block from whichever pieces arrive first instead of waiting for specific ones.
The design has a practical consequence for adoption. Because the channel sits beside the default protocol, a failure or slowdown in the new software should not cut a validator off from the network. That is the general logic of running a parallel route, though the announcement does not describe failure testing in detail.
Ethereum's default gossip layer works by having each node pass new messages to a handful of peers, which pass them on in turn until the data has spread across the network. The method is robust and decentralized, but each hop adds delay, and nodes with poor connections can receive data later than better placed ones. Alternative propagation schemes try to shorten that path or to make delivery less dependent on any single route.
6. What do the speed figures show, and how reliable are they?
In early mainnet measurements, Optimum says blocks arrived 2.42 times faster than they did over GossipSub, the peer-to-peer protocol that Ethereum's consensus layer uses by default. The variance in arrival time was about 7.88 times lower, and users of mump2p received the block first in nine of 10 measurements.
These numbers are self-reported, and the announcement does not explain how they were collected. The same two multiples appeared earlier in a Four Pillars report in August that Optimum sponsored. That report described a test across 30 gateways, carried out while the company was still preparing its mainnet integration. A sponsored study is not an independent confirmation, so outside replication would strengthen the case.
Lower variance is as important as raw speed. A network that delivers blocks at a steady pace is easier for operators to plan around than one with occasional long delays. If the 7.88 times figure holds up in independent testing, it would be the more meaningful part of the result.
7. From private testnet to production
Optimum ran a private testnet with seven operators in June 2025, two months after closing an $11 million seed round led by 1kx. The path from that test to a mainnet deployment with 16 operators took roughly sixteen months.
The company says connecting is free and permissionless. It says integration requires no changes to consensus or execution software and takes one engineer less than an hour. In August, however, Optimum described the network as permissioned for the time being, with operators connecting through an API key that the company issues. The two descriptions point to a gap between the stated goal of open access and the current gatekeeping arrangement.
Optimum also plans to bring the protocol to other high-throughput blockchains after Ethereum. That would extend the same approach to networks where fast propagation is a central concern.
8. What to watch next
The main open questions are about verification and concentration. Independent measurements of block arrival times would show whether the 2.42 times and 7.88 times figures reproduce outside Optimum's own testing, and a clear definition of "Total Value Boosted" would make the headline value figure easier to evaluate.
Observers will also track whether other large staking providers follow ether.fi's lead. The 16 operators now account for just over 3% of staked ETH by Optimum's count, so the software still sits well short of broad coverage. Whether that share rises, and whether the permissioned API-key arrangement gives way to genuinely open access, will determine how much weight mump2p carries in Ethereum's staking infrastructure over the coming months.
Beyond measurement, the mandate raises a familiar question about client and software diversity. Ethereum's resilience has long relied on no single implementation or vendor becoming indispensable. A requirement that spreads one company's software across a large share of a major staking provider's validators does not by itself threaten that principle, since the channel runs alongside the default protocol, but it is the kind of dependency that stakers and researchers tend to examine closely as the footprint grows.