From listening to the talk about Ouroboros Hydra tonight and the future of 10K+ transactions per second due to having many simultaneous epochs going (which is incredible for Ada holders!), as Charles noted even if your block is small you quickly get into gigabyte per second of data…so in my mind, the level of server power needed for staking pools is going to escalate exponentially and there simply won’t be that many total pools regardless of how fragmented it is now.
This is no different than Bitcoin which you could mine on your desktop initially vs now you need super high powered custom ASIC machines and no one’s desktop can remotely compete.
In order to expand the speed of the network the server resources required for pools to meet the network criterion will continue to grow and so longer term you are right, I suspect only the largest and best funded/resource rich pools will exist.
I hope Cardano POS dosnt become like btc mining - but In theory and in long term - makes more sense to expect less pools (but bigger) than lots of small ones… but as the protocol is open source - and we will be able to vote for it - maybe we will change it to not be so centralized around a small group of giant pools.
Marry me…
Sure baby, lets do it 
I know we’re all speculating until the formal announcement clarifies how the stake pools will precisely work, but I’m not so sure that Cardano will inevitably degenerate into the ASIC-based feudalism that exists with Bitcoin today.
Based on the limited information that we have so far, my (optimistic) interpretation of Ouroboros Hydra is that it will more closely resemble the BitTorrent Protocol, which is inherently designed to asynchronously distribute the workload across all the nodes on the network. And like the performance of BitTorrent, the performance of Cardano is expected to increase as the number of nodes on the network increases.
In this context, my understanding is that the network will include full nodes, lightweight nodes, and stake pool nodes. And even though there will be many more lightweight nodes vs. full nodes and stake pool nodes, the BitTorrent-like performance will collectively result in the 10k+ transactions/sec (TPS) that Charles has mentioned in his interviews.
The team will also be implementing side-chains, atomic swaps, and other tricks to maximize total TPS. Thus, unless there is some fundamental law of physics that prevents Cardano from performing like BitTorrent, then I think we still have good reasons to be optimistic that Bitcoin feudalism will not be replaced by Cardano feudalism.
So, until more contradictory evidence is presented to us, I will continue quixotically believing that Cardano will usher in a 21st Century Glorious Revolution, which will enable us to overthrow King Bitcoin and its ASIC minions so that all of humanity can enjoy true and sustainable economic and political freedom.
Thanks @MegaWind
my question, indeed is was about the basics behind a practical stake pools protocol…
The security concern is quite real if you have to keep an exposed node. I guess some of sub-staking can be implemented at least in smart contract or even natively in protocol, but in the end 1 node must be up to do the TX&broadcast right?
Instead of DDoS protection I was thinking maybe you can consider the cost of running a few nodes, not sure how much cost, but real ips run for 5$ on VPS last time I checked. You can think of some rotation schema every 30sec or longer. Just some thoughts. I am new to the whole new PoS/DPoS but not p2p in general 
Hi Pip010,
1 - You have an excellent idea re: rotation. It’s not something I would want to do unless there was an active attack, but we have some unique options with the infrastructure we are using - specifically floating IPs. Thus, we can rapidly swap an IP for a node,and immediately re-register with Cardano network, and in effect the attacker would have to come ‘find us’ again.
It’s not a perm defense,but it would make us a much, much more evasive target for sure. Thanks for the idea!
2 - As a general update re: DDoS defense - I have another meeting with the AI DDoS firm tomorrow. They’ve been reviewing the Cardano docs and will make a proposal on how to best setup. I’m kind of hoping they will put a system on one of our nodes and then it can train/learn now, to be ready for actual staking.
For reference, I did find that IOHK has put some thought into DDoS defense - they list it as a work in progress but note that their main defenses are transaction fees and also relying on ban lists for bad nodes…which they admit is not the greatest, but I’m happy they are thinking about it.
Thanks again for the rotation idea - I’m going to check into whether we would could automate with the floating IP aspect since we have nodes in three different geographic locations, and/or could simply keep ‘spares’ ready for each node to swap at the first inkling of a flood type attack.
Thanks for a nice article MegaWind. I do like seeing the level of detail provided on progress, and the level of effort you are willing to put into the Cardano project. Interesting to see that you are monitor the servers with AI that determines if they are healthy and when issues may arise.
Hi Gordondevoe!
Thanks for the post - to answer your questions:
1- We are using a specialized service provider. They are easily 2x as fast as AWS. Specs on our nodes are: 6 CPU / 8 GB Ram / 200GB Mesh SSD drives.
The above is overkill relatively speaking as they run with about 1% CPU on average…but I wanted to ensure we had way more resources available since we don’t know what will change when live staking starts, and secondly, to build from source, you do need 8GB Ram and > 2 CPU.
2 - Re: power - should be mostly bandwidth and then CPU. I doubt GPU will ever come into play for Ouroboros, any version,as GPU is really only needed for PoW mining vs. we’ll always be Proof of Stake. In O Hydra, it will become a DAG with multiple epochs going at once and then you could be looking at a max of up to 1GB/minute so bandwidth and CPU processing would become the main issues.
Hope you are doing well!
Hi Rickymac,
Thanks for the kind feedback! We are definitely working hard to make sure we have as professional setup as possible.
We’ve added a fourth server now running an explorer node, which will let us start building a couple tools like Ada address verification service for people to use to check addresses before sending…that should go up on AdaView.io in the near future.
Hope you are doing well!
There is no difference AFAIK between delegating and direct. You will be slot leader as often as you win the lottery. Simples.
Lottery is the inside mechanism of Ouroboros, not the rules that will decide you will get rewards! The rewards for staking will be proportional to what you stake! If there will be a lottery than some will get rewards (lucky) some not (unlucky?)
Not sure what you are saying. From my understanding the slot leader makes a block and gets the reward for that. So someone who stakes a lot will be slot leader more often that someone who stakes a little.
Yes. he will get ellected proportionaly random to his stake )(
Can someone give me an approximate cost per epoch of running a pool according to Cardano’s specs? Does the network have a cap on how much you charge or its a free for all?