- The Ethereum ecosystem relies heavily on cloud computing like that provided by Amazon.
- An overreliance on it could cause millions in losses for those staking Ether and impact usability.
DeFi’s main selling point is its decentralisation — no bosses, no hierarchies, and accessible to all. No other blockchain offers that promise like Ethereum.
But an overreliance on cloud computing servers including Amazon threatens that decentralisation — creating weak points that could render Ethereum unusable if such servers go offline.
“What good is it to have a diversity of people running nodes and a geographically distributed network if in the end it’s all running with the same cloud provider?” Sajida Zouarhi, a product manager at Blocknative, an Ethereum data analytics firm, told DL News.
Many companies that provide vital Ethereum infrastructure, such as ConsenSys, rely on virtual servers rented from cloud computing providers like Amazon Web Services.
And a large portion of Ethereum’s execution nodes — software instances responsible for producing new blocks on the network — are also hosted virtually on cloud servers.
If too much Ethereum infrastructure is hosted via one provider or at one data centre, any issues — such as power outages or other technical issues — threaten to knock large chunks of the network and the technology built on top of it offline.
It also gives cloud service providers the power to hamstring the blockchain networks which rely on them.
“Ethereum could do better,” StrategicHash, a pseudonymous blockchain researcher, told DL News.
He cited data from ethernodes.org, which shows 2,439 of Ethereum’s 6,810 nodes are hosted via AWS — about 35%. “Ideally we want one service provider to hold a maximum of 25%.”
Of these 2,439 AWS-hosted nodes, around 1,395 are located in a single data centre in Ashburn, Virginia.
Still, cloud hosting — in particular through AWS — is popular. It is often cost-effective for smaller organisations that can’t afford to purchase and maintain hardware themselves.
For those running Ethereum nodes, cloud services are also seen as more reliable, as the large data centres where they’re hosted have access to more stable electricity and internet connections.
An AWS outage is unlikely to bring Ethereum to a standstill — other nodes hosted elsewhere would continue to process transactions.
However, StrategicHash said, an outage could still cause a lot of damage.
He explained that those staking Ether through nodes hosted on AWS would get slashed due to inactivity, resulting in thousands, if not millions of dollars of losses.
Slashing refers to taking away some Ether from nodes who either act maliciously, or don’t do their job correctly, like being offline when they should be checking transactions.
An outage would also reduce the number of nodes on Ethereum making the network temporarily more centralised and susceptible to attack.
But while Ethereum will likely keep chugging along, an overreliance on AWS hosting could still be a big problem for the blockchain’s users.
‘An AWS incident can have a vast impact on user-facing applications such as hosted frontends or crypto wallets’— Pascal Caversaccio
“I’m not that much afraid of any consensus failures due to cloud hosting but rather on its impact on the user,” Pascal Caversaccio, a blockchain security researcher, told DL News.
“An AWS incident can have a vast impact on user-facing applications such as hosted frontends or crypto wallets,” Caversaccio said.
He noted MetaMask, the most popular Ethereum wallet with over 21 million monthly users, relies on software hosted on AWS to submit transactions to Ethereum.
If the AWS servers where this software is hosted go down, MetaMask effectively stops working.
Such a situation played out last year when Infura, MetaMask’s default remote procedure call provider, which lets the wallet’s users interact with blockchains, experienced an hours-long outage.
While the problem can be fixed by switching from MetaMask’s default RPC to a third-party provider, most users won’t have the technical ability to do so.
“It’s important to actively educate people what to do in such situations,” Caversaccio said.
Even DeFi protocols — which sell themselves on their ability to process transactions around-the-clock — are not immune. Some rely on AWS.
In December 2021, an AWS outage caused decentralised exchange dYdX to stop processing transactions. In a tweet, dYdX said “there are still some parts of the exchange that rely on centralised services,” mentioning AWS specifically.
But the data shows Ethereum is becoming less reliant on cloud services. In 2020, over 70% of execution nodes were hosted virtually.
Since then, that number has decreased to around 60% as more chose to run nodes from their homes to make the network more decentralised and resilient.
However, despite the move away from cloud hosting, Amazon Web Services has become more dominant, and now accounts for a higher proportion of hosted Ethereum nodes than it did before.
Additionally, there are other ways to benefit from the perks of remote cloud servers while mitigating some of the risks.
“A possible solution is to encourage hosting on bare metal rather than cloud providers,” StrategicHash said.
Bare metal refers to physical computer servers that are used by one tenant only, as opposed to cloud servers which share hardware with several other tenants.
“Bare Metal is cheaper, faster and is unlikely to go down all at once,” StrategicHash said.
Do you have a tip or information on how DeFi uses cloud computing or another story? Reach out to me at email@example.com.