The Ethereum scaling story tends to center around layer-2 rollups, but what about the EVM itself? Does the virtual machine suffer from limitations that are inherent to the execution layer of the system, or can developers overcome them?
While rollups take the computation load off the Ethereum mainchain, Blockworks co-founder Michael Ippolito wondered how the computation that occurs directly on Ethereum could be more “performant.”
“Do rollups solve all of our scalability challenges, or is there still work to be done on the execution layer?” he asked during the Permissionless II conference in Austin, Texas.
Monad Labs developer relations engineer Kevin Galler replied: “A lot of the current limitations of the EVM we know today are actually more of a side effect of how the backend computer of this EVM has been constructed up until now,” but are not inherent to the interface or the bytecode.
“You can actually design parallelism and stuff into this,” he said, acknowledging that “there’s stuff you have to work around to maintain backwards compatibility.”
“There’s like, 20 years of high performance computing research on how to do these things,” he said, “so, it’s a very solvable problem.”
Offchain Labs senior software engineer Raul Jordan admitted that the current implementation of the EVM is “fairly slow for many reasons,” adding “there are many things we can do to power that up.”
Jordan noted that developers are already accustomed to being able to build DeFi primitives and other “really cool applications” with the EVM as it stands today, but wondered “how can we give them superpowers to build something greater than they have before?”
“How can we empower them?” he asked. “Should we make the EVM faster, or can we introduce new constructs that allow them to do things that weren’t possible before?”
Can’t cheat physics
Somani expressed concerns with the possibility of an excess of layer-2 protocols. “The issue with the single threaded EVM…is that it results in a lot of rollups.”
“There’s people suggesting there should be millions of rollups or something,” Somani continued. “That’s going to lead to a lot of issues with liquidity fragmentation and end-to-user experience.”
“Sufficiently advanced virtual machines” have local fee markets, Somani said, “so that if there’s a lot of activity on one app, it doesn’t congest all those.”
“So bringing that [virtual machine] to Ethereum can lead to this highly parallelized [layer-2] that doesn’t really need all these other robots to exist,” he said.
“You can’t cheat physics,” Galler interjected. “Bandwidth is fundamentally what limits how fast a lot of these blockchains can run.”
Reflecting on scaling challenges, Jordan referred to the example of Arbitrum Stylus, noting that the product allows developers to write code in common programming languages like Rust and C++, compiling to a special instruction format called WebAssembly. “You can actually execute this side by side with EVM smart contracts.”
“So we ended up realizing that the EVM is not the ceiling,” he said, “it’s actually the floor.”
“You can do a lot more than you thought was possible while still maintaining the same guarantees.”
Scalability at the execution layer “has a long way to go,” Jordan added, “and we’re trying to get there.”