cryptonist
A printed circuit board photographed flat, its traces and mounted components in rows

Explainers

What Solana is, and what its on-chain data shows

Solana is a blockchain built to process transactions quickly on one shared ledger, rather than spreading the work across separate rollups. It keeps time with a cryptographic clock called proof of history and agrees on blocks through votes weighted by staked SOL, and on seven occasions documented in reports on its own site it has stopped producing or finalising blocks, or slowed badly, since 2020. Its DeFi deposits rank third among the chains this site tracks, and, as on Ethereum, the largest category on its data page is one the chain total leaves out.

What the chain is for

Anza’s validator documentation says Solana operates as a single global state machine. What lets one ledger carry a large volume is that work can be done in parallel. A Solana transaction states in advance which pieces of state it will read and which it will write, as the project’s April 2022 outage report explains: reads can overlap and run in parallel, writes to the same state cannot, so transactions that do not write to the same accounts can be processed side by side. A chain that executes transactions strictly one after another cannot.

It differs from Ethereum below the interface too. Solana signs with a different curve and formats its addresses differently, which is why a wallet built for one does not automatically work on the other; why wallets are chain-specific explains the engineering.

The network that the project’s reports call Mainnet Beta started on 16 March 2020: the chain’s own record gives its first block a timestamp of 14:29 UTC that day, which a public node returns for the time of block 0. The Solana Foundation describes itself as a non-profit foundation based in Zug, Switzerland, dedicated to the decentralisation, adoption and security of the ecosystem. Validators run one of several independent clients: Agave, which its documentation describes as a fork of the original validator once maintained by Solana Labs and now developed by the engineering team at Anza, and Firedancer, a client written from scratch, which its documentation says brings client diversity to the network. Changes to the protocol are proposed as Solana Improvement Documents in a public repository.

How consensus works

Two mechanisms do different jobs. Proof of history keeps time; proof of stake decides who produces blocks and whose votes count.

Proof of history. Solana’s terminology page defines it as a stack of proofs, each showing that some data existed before the proof was created and that a precise duration of time had passed. Like a verifiable delay function, which the same page describes as taking a fixed amount of time to run, it can be verified in less time than it took to produce. The design document for Tower BFT, Solana’s voting algorithm, calls it the cluster’s source of time. Time is counted in slots, and each slot has one designated leader that may produce a block in it.

The leader schedule. Which validator leads which slot is set an epoch in advance. Anza’s leader-rotation page describes the method: validators take the staked accounts that have voted recently, order them by stake weight, and use a pseudo-random seed to draw a stake-weighted schedule. Every validator computes the same schedule, so each can reject a block not signed by the expected leader. Only one leader produces at a time, which a malicious leader could abuse by censoring transactions, and rotating the role is how the design limits that influence.

Tower BFT. Validators vote for the fork they consider heaviest, and each vote carries a lockout: a period, measured in slots, during which the validator may not vote for a conflicting fork. The design document requires the cost of breaking an older vote to rise faster than linearly with its age, so the longer a validator has stood behind a fork, the more it costs to leave it. The terminology page defines finality as the point at which nodes representing two-thirds of the stake have a common root.

Stake. SOL holders who do not run a validator delegate to one. Solana’s staking documentation says that the more stake delegated to a validator, the more often it is chosen to write transactions to the ledger, and that validators take a commission from the rewards they share with delegators. It also says, in terms, that there is currently no in-protocol implementation of slashing. That is a large difference from Ethereum, where the protocol itself destroys stake for provable misbehaviour.

A replacement has been proposed. SIMD-0326, still marked Review in the repository, proposes replacing the consensus protocol based on proof of history and Tower BFT with the voting component of a new design, Alpenglow. Its motivation section says Tower BFT has a finality time of 12.8 seconds and lacks a security proof. Anza’s documentation calls Alpenglow upcoming, expects it to activate in the Agave v4.3 release cycle and says it targets finality in roughly 150 milliseconds. Until it activates, the mechanism above is the one running.

What SOL does, and how fees work

Every transaction pays a fee in SOL, and Solana’s fee documentation splits it in two. The base fee is 5,000 lamports per signature, a lamport being 0.000000001 SOL, and it pays validators for the work of verifying signatures; half of it is burned and half goes to the validator that produced the block. The optional prioritisation fee is the compute-unit price the sender sets multiplied by the compute units requested, and all of it goes to the validator. It raises the chance that the leader schedules the transaction ahead of competing ones.

The contrast with Ethereum’s model is in which part moves. On Ethereum the protocol raises the base fee for everyone when blocks are full, as what gas is describes. On Solana the base fee is fixed per signature, and contention is handled through priority fees. The April 2022 report gave the reasoning: congestion on Solana is less often a fight over the whole block than over one piece of state, such as a single minting contract, so fee pressure should fall on the transactions that want that state and not on everyone.

SOL’s other use is stake. Delegated SOL decides how often a validator leads and how much its votes weigh, and some of it is staked through services that issue a token for it, the mechanism liquid staking explains. Those services are the largest category on Solana’s data page.

The outages

Solana’s own site carries reports on the failures below. Each of the seven ended with validator operators coordinating a restart of the network; none was resolved by the protocol on its own. The dates and times are the reports’ own.

DateWhat happenedCause, per the reportRestored
4 Dec 2020New block confirmations haltedA validator running two instances sent different blocks for one slot, splitting the network into three minority partitions that a known bug in block repair left unable to reconcileRestart within 6 hours, across 393 validators
14 Sep 2021Network offline for 17 hoursBot transactions around a token sale flooded a queue whose memory grew without limit, crashing validatorsRestart from the last confirmed slot, which needs 80% of active stake
30 Apr 2022Block production stopped at about 20:30 UTCBots competing for an NFT mint sent 6 million transactions a second; validators ran out of memoryResumed at 03:30 UTC on 1 May
1 Jun 2022Block production stopped at about 16:30 UTCA runtime bug let a failed durable-nonce transaction be processed twice, splitting validatorsResumed at 21:00 UTC, with durable nonces disabled
30 Sep 2022Cluster halted at about 22:41 UTCDuplicate blocks from a malfunctioning hot-spare node hit a bug in fork-selection logicResumed at about 06:57 UTC the next day
25 Feb 2023Finalisation slowed from 05:46 UTC; leaders went into vote-only modeAn unusually large block, re-forwarded by block-forwarding services, overwhelmed block propagationManual restart with a software downgrade; normal finalisation from 01:28 UTC on 26 Feb
6 Feb 2024Block finalisation halted at 09:53 UTCA bug in how the validator cached compiled programs deployed with a legacy loaderConsensus resumed at 14:55 UTC, about five hours later

The causes vary more than the pattern. Two began with load, bots competing for the same thing at the same moment; two began with one operator’s node producing duplicate blocks, which then met a bug; one began with a malfunctioning validator’s oversized block, which block-forwarding services then amplified; and two were bugs in the validator software itself. The 2021 and 2020 reports both stress that no funds were lost or at risk. A halted chain does not lose balances, but nothing can move while it is down, and anything with a deadline, a loan near liquidation or an order waiting to fill, waits too.

The most recent outage report on solana.com covers February 2024. The same site has since published two post-mortems of a different kind. Security researchers reported flaws in the ZK ElGamal Proof program, which verifies the zero-knowledge proofs behind confidential token transfers: the first in April 2025, patched across validators within days, and a second in June 2025, after which confidential transfers in the Token-2022 program were disabled pending further audit. The first report says there is no known exploit of the vulnerability.

DefiLlama’s hacks record, as committed to this site, lists 61 incidents tagged Solana between 2021 and 2026, 36 of them filed against DeFi protocols and 11 against centralised exchanges; the Solana hacks page breaks them down by year, counting a multi-chain incident in full on each chain it touched. These are failures of applications and services on the chain, which how crypto gets hacked classifies across the whole record.

What the on-chain data shows

As of 22 November 2026, per DefiLlama, Solana had $4.6B of total value locked, down 21.3% over 30 days, which ranks it third of the 121 chains this site tracks (its data page). DefiLlama lists 426 protocols deployed on it, and $13.0B of stablecoin supply circulates there.

The total value locked figure is the value of assets deposited in DeFi protocols on the chain, at the snapshot date’s prices. The chain figure leaves out liquid staking, bridges and deposits DefiLlama flags as double-counted, and native staking is not counted at all: DefiLlama’s FAQ says it counts no chain’s staked assets, SOL among its examples. What TVL measures sets out why. The page’s list of the fifteen largest protocols shows each at its own figure, so it includes what the headline leaves out:

ProtocolCategoryTVL on Solana
Sanctum Validator LSTsLiquid Staking$1.3B
Kamino LendLending$1.1B
Jupiter LendLending$953.9M
Raydium AMMDexs$780.7M
Binance Staked SOLLiquid Staking$723.0M

Five of the fifteen are liquid-staking entries, together $3.4B, and the fifteen add up to $8.4B, against the chain’s $4.6B (both computed). The list is not a breakdown of the headline and is not meant to add up to it.

The two change figures say something about timing. TVL fell 21.3% over 30 days but only 0.3% over the last seven, so almost all of the month’s fall came before the last seven days (computed from the two). The snapshot carries dollar values only, and a dollar figure falls when deposits leave and when the prices of the deposited tokens fall, so it cannot say which drove the move.

Stablecoin supply on the chain, $13.0B, is 2.8 times its total value locked (computed). Solana’s mix is also different from the other large chains. In this site’s stablecoin table, built from a separate DefiLlama feed, USD Coin holds $6.8B of the $13.3B placed on Solana, 51.0%, and Tether $2.0B, 15.3% (computed); on Ethereum, Tron and BSC, Tether is the larger. Where stablecoin supply lives reads the same table chain by chain.

What the page cannot tell you

How reliable the chain is. Deposits measure what people have committed to the chain’s applications. They say nothing about uptime, and a figure taken on a day the network was running looks the same as any other. The outage reports are the record for that.

How much SOL is staked. Native staking sits outside the TVL rules, and the liquid-staking entries cover only the part of staking that issues a token.

Who uses it, and how much. A deposit total is not a count of users or of transactions, and on a chain built around throughput, transactions are much of the point.