Thread by @_KtorZ_

Some months ago, I was asked what I thought would be #Cardano "next big thing".

At the time, we only had a vague idea of what Leios was. The recent CIP from Duncan & al has shed some more light on it.

Ouroboros Leios is definitely the most promising L1 upgrade to come

👇

The idea behind Leios is (surprisingly) simple. So let's try to make sense of it.

Cardano is a distributed system, where nodes reach consensus by sharing blocks according to a schedule for which they're elected. The more stake they own, the higher the chances of being elected.

Blocks are produced about every 20s, and are propagated to the rest of the network as fast as possible.

If we zoom in a bit on a node and analyze what a node does, we see that most of the time, a node isn't doing much. It is actually idle for a large part of these 20s.

This is partly due to how networking work. Sending data across the planet takes time, and while the data is being sent from a point to another, both ends aren't actually doing anything but waiting. Serializing and deserializing from and to the wire only account for a small time.

Similarly, while a node is serializing data and sending out data to its peers, it is usually not actively validating blocks, because the block throughput isn't usually sufficiently high to keep the node busy at all time. The CPU usage only increases in short spikes.

If you were at the #CardanoSummit2022, you might have seen the pool setup by @C1cADA_Markus on the 3rd floor and observed that the node would only have short spikes of activity every 20s or so.

Said differently, resources are currently under utilized in the node.

Fundamentally, this is because the Cardano blockchain is constructed and validated sequentially (like a chain!)

The idea behind Leios is about utilizing those idle resources, and making the construction of the chain more parallel.

It is about harnessing the true power of UTxO.

Leios proposes to separate the construction of the chain into three new types of blocks, produced at different rates and with different purposes:

- Input blocks (most frequent)
- Endorsement blocks (high confidence agreements)
- Ranking blocks (true consensus)

Ranking blocks are similar to current consensus blocks. They are produced sequentially and serve as 'checkpoints' allowing everyone to regularly agree on some global state.

Yet instead of containing plain transactions, they would only contain references to endorsement blocks.

Endorsement blocks are ways for the chain to reach an agreement, with a relatively high degree of confidence. With them, Leios proposes to introduce endorsement certificates which inform nodes about how many peers (or even, stake) are approving certain events.

For a block producer, it means that at a given point in time, they can now how much of the network stake has approved (or, endorsed) some given transactions without actually having to download them.

This allows nodes to optimistically make consensus decisions at low cost.

Finally, Input blocks are bundle of transactions. They are the actual data payload containing information. They are meant to be produced at a fast rate and validated concurrently. When spread over multiple machines, it means validation happen with some degree of parallelism.

This approach works because of one key assumption/observation: most transactions are actually 'concurrent'.

Or put simply, most transactions deal with different, non conflicting, parts of the ledger (a.k.a. UTxO entries).

A UTxO model makes it relatively easy to assess the concurrent nature of two transactions; for there's no side effect.

It suffices to look at transactions' respective inputs and outputs to know whether they are conflicting with one another. There's no hidden consequences.

This makes Ouroboros Leios a fundamentally *optimistic* consensus algorithm. And because this hypothesis is so important to the system, Leios also includes in its design a way to degrade into a more sequential (pessimistic) mode, which operates very much like Ouroboros Praos.

There are other details which are related to security and usability considerations, as well as still open questions about rewards and incentives in this new scheme.

I'll recommend anyone interested to read the current proposal:
github.com/cardano-founda…

It is very digestable 😊

More, to manage a bit expectations, keep in mind that Leios is undoubtedly far away from now.

This is a massive chunk of work and probably the most ambitious upgrade Cardano has ever attempted (feels like we say this every 6 month?)

Meanwhile, L2 is also booming 😏

Finally, I'd like to point another thread from @zygomeb who's going into some more details about how the VRF leader election is going to work on Leios and gives another perspective on this proposal. A good read as well!

https://twitter.com/zygomeb/status/1595470922621894661

Keep buidling, stay awesome 💛