Twitter community and fellow Hbarbarian, If you haven't voted yet,

here are 6 reasons why you should vote Hedera vs Elrond in the link below.
Let crypto twitter realise the power of the hashgraph.

1) Throughput:
Hashgraph is able to achieve 100,000 TPS/sec/shard (Artificially throttled to 10,000+ currently)

Elrond's pBFT-like Algorithm approaches, according to whitepaper, peak VISA level (55,000tps) with 16 shards, that's a max theoretical limit of 3500tps/shard.

2/ Latency:
Hedera achieves deterministic finality within 3/7 sec

Elrond has a round block m time of 6 sec, but still needs to wait at least block m+1 to be notarized by the metachain, that's another 6 sec. But what about consensus time on the metachain? 🤔

3/ Decentralization:
Hedera has 39 nodes to bootstrap the network but will gradually scale to 100s of nodes within 1 shard. All nodes are equal, all nodes participate, all are rewarded.

Elrond, every 6 sec, randomly picks 63 nodes with the highest stake in a shard of 800 nodes. Only the influencial ones vote, the remaining nodes don't get to play.

4/ Synchronous vs aBFT:
Hedera is asynchronous, ie doesn't require the notion of time coordination (a drum beat) to progress efficiently. Many nodes can go offline, delay messages and consensus will still be reached.

Elrond pBFT is synchronous. Any faulty/malicious node could cause latency issues in reaching consensus.

5/ Leaderless vs Leader-Based:
Hedera is leaderless, all nodes gossip transactions at the same time. No single point of failure. Full decentralization

Elrond uses a VRF function (which could be rigged) to randomly elect leaders. A leader is a single point of failure and could be susceptible to DDOS attacks.

6/ Metachain vs Equal Shards:
Hedera is aBFT and reaches deterministic finality within secs allowing shards to directly communicate.

Elrond's Metachain is a central hub to coordinate intershard communication and requires its 400 nodes to reach consensus adding complication and extra latency.
Listen to Dr Baird on how efficient  sharding ought to work 1:02:15

https://twitter.com/DanyEid_/status/1510347555913281540?t=LBa2wamt5z1yNt5kgsaKAA&s=19

3) sorry where you have that information that will scale to 100s of nodes within 1 shard. That would be absurd amount of nodes per shard with high latency and reward fee from each transaction would have to be shared split within 100s nodes. It would never work like that from both economic and technical point of view.

Great post as always 👍🏻

Check Leemon's Videos. Always mentions hundreds of nodes within a Shard. The white paper mentions 128 tested

Thank you. Nice investigation work yourself 😉

For testing maybe, he mentioned in Town Hall even 26 is secure and there was clear latency goes up more after certain amount. There is no need even for that amount.

A higher number could be required to mitigate the liveness issue. You dont want to risk many staked nodes going offline all at once. This could jeopardise the safety of a shard. But yeah, maybe a hundred or 2 could be the optimum between performance (Latency as you put it) and safety.

The 14 to 16 is good number/ shard, but powerful with as much TPS as possible per shard and not run by community. You have example what happens with Elrond when shards are run by community. TPS alone is one thing and how network achieve it is another. Fragmentation is never good, and what if single app requires more TPS than single shard, then need to monitor another shard or another and stream history from multiple. Hedera did shard testing in close environment and not using community nodes for that, so discussion how many TPS shard can take in Hedera is irrelevant at the moment, if they solution for scalling is using community nodes. There never been network to date that achieved scalling using community nodes. Elrond, if Visa wanted to use it and they have 16 shards to cover just their level, means 16x increased risk as each shard is it's own environment with each own risk for any network. Also history to be monitored and recorded from 16 places instead of one.

They call them community nodes but these are KYCed, approved and are expected to adhere to the minimum specifications required to host a node.

Elrond has a feature whereas they score reputations for nodes (not sure of the details) but a node for instance could be "jailed" for a period of time, for example if it fails to maintain over 95% online presence, or if it fails to sync or vote, and this could even reflect over the nodes returns, since if a node is offline it also reflects on the reduced profits of the wallets proxy staking to it, which will result obviously in them unstaking and staking with more reliable nodes. This will eliminate faulty nodes and a reputation system would always highlight the top X reliable ones for example

I doubt they have slower protocol per say, that may be just problem when you run community nodes you get what you get as TPS per shard. I looked on their website, they will scale to certain level and that's it, otherwise found nothing interesting or unique what they doing. Just another crypto out there.

They were probably the first to implement some sort of sharding but yeah you're right it's the way sharding is implemented that matters, and their model could hit a bottleneck. I mean you still have to coordinate all shards through the metachain anyway so this could be a bottleneck. In all cases, let's see how Leemon solves this issue. Let us hit those 10,000 tps/ shard first 😊 And with that, I wish you a good night Michal 🙏

Have a good night too. Sharding everyone will do including Bitcoin their model is bit different. We see what actual usage TPS numbers numbers we get end of 2022 and what the throttle will be for different networks out there.