Ethereum 2.0 Developer: PoS Merger Plan May Be Implemented Early Next Year at the Latest

Note: The original author is ConsenSys researcher and Ethereum 2.0 developer Ben Edgington. He expects the merger phase of Ethereum 1.0 and 2.0 to be completed at the end of 2021 or early 2022, which means that the Ethereum system will switch from the proof-of-work (PoW) consensus mechanism to proof-of-stake (PoS).

Beacon Chain

There are currently over 132,000 active validators, with 4.2 million ETH staked (worth nearly $16 billion as of now), and the rate of new deposits has increased recently.

Since the minor shock two weeks ago, everything has been going smoothly. Here is a detailed review of the incident from Prysmatic Labs.

Unfortunately, after two full months of operation without any slashing, three isolated incidents occurred in the last month, and I have no information as to what caused them.

Staking

One thing I really enjoyed was watching Adrian Sutton sync Teku from scratch in about a minute. The demo downloaded the initial state directly from Infura, although you can use any known synced beacon node or a downloaded file containing the state. This is a game changer for staking services and individual stakers. When syncing a beacon node is so fast (seconds instead of hours), you can basically forget about maintaining persistent storage and redundant setups.

If you want to check how well your validator performs its validation duties, then take a look at Paul Hauner's spreadsheet tool‌ and instructions. He provides more insights than the single "valid" number provided by Beaconcha.in.

Are you aware of Secret Shared Validators? If not, now is a good time to catch up. Alon Muroch has written about the recent SSV Phase 1 testing round.

Rayonism

Rayonism‌, also known as a Russian abstract art style, is an ongoing effort that is being conducted in parallel with the Eth1/Eth2 merged testnet and sharding R&D. It emerged at the EthGlobal Ethereum Scaling Hackathon, but is sure to expand beyond that.

Lukasz Rozmej from the Nethermind team put together a very nice video tutorial showing how to set up a merged testnet using Nethermind for the Eth1/execution side and Teku for the Eth2/consensus side.

The big news is that the first developer testnet, Steklo, was live for just a few hours on Friday, April 30th.

Steklo means "glass" in Russian, which means it will be vulnerable, and it proved to be so. To be fair, all teams infiltrated the network without any test vectors prepared, so this was the first blind attempt. Decrypt wrote an article about Steklo, and there were twelve client combinations in total: Nethermind, Besu, and Geth/Catalyst were paired with Teku, Nimbus, Lighthouse, and Prysm respectively.

Lukasz provided a summary on the Rayonism Discord channel:

1. Lighthouse had consensus issues and forked from the beginning (state root issue), but on this fork all execution engines worked perfectly fine.

2. Prysm encountered an issue (general or consensus issue, no confirmations) and got stuck at the genesis stage;

3. Sometimes Besu or Teku didn't agree with the other clients, but eventually they reached consensus. When split, the split was into sets (Teku-Geth, Teku-Nethermind) and (Teku-Besu, Nimbus-Geth, Nimbus-Besu, Nimbus-Nethermind)

4. Nimbus talking to Nethermind probably had some issues with the transport layer on the Nimbus side, but it required more investigation from both the Nimbus and Nethermind teams and eventually it worked most of the time - probably with some restarts.

In short, a lot happened, and a lot of things went well, and overall the whole exercise was very encouraging. After some repairs, Lighthouse later reached consensus with Teku and Nimbus.

Some common test vectors have been prepared over the last week so that clients can debug individually before coming together again. The plan is to have a longer-lived merged developer network in place early next week.

Altair

Altair is a relatively minor beacon chain update scheduled for release mid-year.

According to recent developer calls, the various client teams are making fairly good progress implementing the Altair spec. We took this opportunity to sketch out a (non-committal) timeline:

1. The Altair spec freeze will occur around May 21st.

2. A short-lived joint test network was launched in early June (not testing fork conversions, only testing the Altair specification)

3. Try to fork the current testnet before the end of June. I guess Pyrmont will be the first one, so it's best to migrate to Prater as soon as possible to avoid instability.

4. Deployed to the beacon chain in late July or early August.

There are still some things that need to be worked out before freezing the spec. In particular, how to handle this issue (losing an epoch of rewards during the upgrade).

merge

Currently, most of the merger activity is focused on Rayonism.

Still, there is a lot of work to be done after Rayonism. Mikhail Kalinin summarized the public conversation in a thread on Rayonism’s Discord channel:

First, the transition process (aka docking). We need to figure it out, write code and try it out in test/dev networks. Second, the sync algorithm; design work is already underway, then implementation and testing. Block proposal optimization techniques, communication channel protocol, whether JSON-RPC or REST; JSON-RPC modifications for users (adding a finalization block method). Writing specific EIPs that describe execution layer modifications. Fixing BLOCKHASH randomness issues. Multiple rounds of various tests. Chain tools, find evidence that they are useful. Infrastructure updates, the main question is what the block explorer will look like.

We have a lot of things to do, but the highly optimistic view is that the merger will be completed by the end of the year, the more cautious view is in the first quarter of 2022, but no one expects it to be later than that.

tool

My former colleague Mostafa Farghaly posted Kotal:

Kotal is a blockchain deployer that makes it super easy to deploy highly available, self-managed, self-healing blockchain infrastructure (network, nodes, storage clusters, ...) on any cloud.

It includes support for Teku, Nimbus, Lighthouse, and Prysm nodes, as well as IPFS and other Web 3 support.

Original article: https://hackmd.io/@benjaminion/eth2_news/https%3A%2F%2Fhackmd.io%2F%40benjaminion%2Fwnie2_210508

By Ben Edgington