Skip to content

The CCS Interview

Casper Labs CEO says TPS is a flawed metric as 22 validators join testnet

Mrinal Manohar, CEO of Casper Labs, spoke to Crypto Coin Show in April 2020, shortly after the project's first public testnet launched on March 31. He explains why Casper puts decentralization ahead of raw throughput, how its provably live CBC Casper variant works, and what developers and validators can expect. He also covers the planned role of the CLX token.

Interview by Ashton Addison Recorded April 2020 5 min read

Key takeaways

  • Casper Labs is building a fully decentralized proof-of-stake blockchain that aims to achieve scalability without compromising security or permissionless principles.
  • The Highway protocol, developed by Casper Labs, is a provably live version of CBC Casper that will be included in their first release.
  • Proof-of-stake systems achieve efficiency gains by eliminating the 90-95 percent of processing power used for random number generation in proof-of-work systems.
  • Casper Labs testnet launched on March 31st with 22 validators and the team expects the mainnet to be code complete in fall with a launch planned for that year.
  • Casper Labs operates as a completely open and permissionless system where anyone can become a validator by staking, with no special classes or requirements.

Decentralization before throughput

Ashton Addison: What is Casper Labs, and why do you see it as the decentralized future?▶ 00:21

Mrinal Manohar: I think the industry has had two big tectonic shifts. The first was Nakamoto consensus, on which Bitcoin is based. Its core innovation is that it's the best form of copy protection the world has ever seen. Many people tried to copy-protect MP3s and other digital goods and never succeeded, and this found a way to do it for something digital, so the digital asset was born. The second was Ethereum adding Turing-complete smart contracts, which made those assets programmable. The next shift, I think, is scalability. Some people think that's just throughput, but I think of it as throughput, security and developer access. Systems aren't scalable unless you get all three, and a lot of projects have compromised, putting true decentralization and permissionless principles aside for the sake of expediency on throughput. I call this the TPS dilemma. I don't think TPS is an excellent metric. A system needs to be fast, but if it isn't decentralized and permissionless, there's no point, in my opinion. Casper Labs is about building a fully decentralized and permissionless system without ever compromising those two things, while providing scale through proof of stake and other engineering. We also take security and developer adoption very seriously, and we think they are core components of scalability too.

Ashton: How does Casper's CBC proof-of-stake algorithm differ from Ethereum's approach?▶ 02:23

Mrinal: It's not that different from where Ethereum wants to get eventually. They do want to get to CBC Casper, which is 100 percent proof of stake. Ethereum is running as a proof-of-work network, so there's a transitionary state, part proof of work and part proof of stake, and then eventually a full switch. The development timeline, last I saw, seems a few years out. Where ours varies is this. Going only off the latest published work, the proposals for CBC Casper at Ethereum were provably safe but not provably live. Provably live means you are guaranteed the system will come to consensus at some point. We published the Highway protocol, a provably live flavor of CBC Casper, and that will be part of our first release. People can read the paper on our website. We're also updating it, so a Highway++ is coming out soon.

Ashton: Can you really balance scalability, security and decentralization together?▶ 04:05

Mrinal: As long as you're not using TPS as your core metric, you can get all three. First, with a pure proof-of-stake system, the first piece of efficiency is this: in proof of work, 90 to 95 percent of your processing power generates random numbers that have no further utility. Proof of stake flips that. No random numbers need to be generated to provide security, so all that processing power can be refocused on useful work, like forming consensus or processing transactions. That's a huge lift right out of the gate. Second, if you do smart things with your virtual machine and your block footprint, you can start scaling to hundreds or thousands of TPS. In tests we're currently getting about 160 to 200 TPS on a single core. If you think of an eight-core machine as standard, the VM is doing around 1,600. We can do that without sacrificing decentralization or security. All the security properties of Nakamoto consensus are pretty much preserved in our proof-of-stake protocol, and we've engineered it from the ground up to be an open, permissionless system.

I call this the TPS dilemma. I don't think TPS is an excellent metric.

Testnet and open validation

Ashton: Where is the Casper Labs testnet now, and what is the timeline to mainnet?▶ 06:10

Mrinal: Our testnet launched on March 31st. We gave ourselves until the end of Q1, so we hit it. We thought we'd have five or six people and keep it small and close, but 22 people signed up, so we have 22 validators running right now. It's our first public, decentralized release, so we're working through some of the issues you discover when you first launch. Overall we're really happy. We have blocks finalizing on the network, most of the validators are staying in sync, and it has given us a basis for a lot of the optimizations we plan to do in the coming months. We think we'll be code complete around the fall, and we're looking to launch this year.

Ashton: Can anyone become a validator, or are there rules validators must follow?▶ 07:19

Mrinal: There are no rules or specifications. Anyone can join. It's a completely open and permissionless system. We don't even have classes, with real validators and super validators. Everyone's exactly the same. You come to the table with your stake and you're able to validate. The only thing I'd say is that we probably have a recommended hardware spec. It won't be super expensive, because this is proof of stake. If you fall behind on the state, your rewards might be throttled a bit because you're not keeping up with the rest of the system. Other than that, there are no special requirements. As long as you have a stake and can keep up with the system, you should be good to go.

Bringing developers on board

Ashton: What tools will Casper offer developers, compared with building on a larger ecosystem like Ethereum?▶ 08:48

Mrinal: First of all, we love this area, and with all due respect, we're standing on the shoulders of giants. None of this is meant as a knock on any project. One thing we've noticed in talking to developer communities is this. There are seven to ten thousand blockchain developers, and most of them know Solidity, but there are 26 million developers worldwide. Penetration of this industry is sub 0.2%. One of the primary reasons is that developers don't know where to start and don't understand the programming languages. So we decided right out of the gate to embrace open programming standards. We'll support Rust and AssemblyScript from the start, and over time any programming language that has the right compilation target, which is about 95 percent of them. People already know these languages, there's tons of documentation, they're well maintained, and they fit into the IDEs and developer environments people already use. There's more. Querying the state of the blockchain has always been an issue. On Ethereum you use third-party tools like Infura. We have a GraphQL interface, so querying state becomes fairly straightforward. We also have continuous integration and continuous deployment, which is very important for businesses. Contracts can be upgraded, and you can decide who can call your contract, from 100 percent open source and callable by anyone to callable only by you. We also have weighted key management and flexible payment code. All of this came from feedback from companies that want to build on blockchain and told us why they aren't.

Everyone's exactly the same. You come to the table with your stake and you're able to validate.

The CLX token

Ashton: What is the primary purpose of the CLX token?▶ 11:39

Mrinal: It's twofold. First, it's the token used to stake as part of our proof-of-stake system, so the underlying security of the system is established by CLX. Second, it's what you use to pay for usage of the system. If you want to deploy a transaction or deploy something from your application, you pay the validator set with the CLX token. It works very similarly to gas on the Ethereum network, except it's also a staking token.

Ashton: How will CLX be distributed, and how do validators get tokens to stake?▶ 12:26

Mrinal: The token will only be released after the network goes live. We're not going to do any interim token. Once we've gone live, it will be listed, and people can acquire it either to pay for computation on the network or to become a validator. The initial validator set can acquire tokens privately, but only for validation.

This interview has been edited for length and clarity. Recorded April 2020.

Mrinal Manohar
Mrinal ManoharMrinal Manohar is the CEO of CasperLabs, a blockchain protocol company. CasperLabs develops a scalable, secure and decentralized blockchain protocol powered by the CBC Casper Proof of Stake consensus algorithm. Full profile →
Share this interviewXLinkedIn
▶ Watching the interview