seems the stress test ended? down to 1.15 tx per second (from way above 2.5) https://tradeblock.com/blockchain

This test is very good on the live network, it showed us some interesting phenomenon and potential area to improve:

1. The cost for transaction is decided on the number of transactions, not how much bitcoin you send, so it will always benefit those who do large transactions, and people should do large transactions as much as possible, to make it economical

2. Larger blocksize like 20MB can not prevent this kind of attack/spam, it will just make it worse. Because carrying out such an attack is always feasible regardless of the blocksize: Spammers can send out 1 million transactions with very little fee, and the cost will be only 20x of this time, but the resulted congestion will be much more severe

3. Mempool will be overflow if the attacker just send millions of transactions, so there should be a mechanism to remove the transactions from the mempool if they are more than xx hours old

4. There is no universal solution to deal with the attack, it will be most efficient if the miners can find the source of the attack and block it, but if the attacker use a large network it will be very difficult to do that. Or, miners could register large exchanges (they typically have high transaction frequency), and only accept frequent transactions from those exchanges

Could the use of sidechains solve the prolem?

But then the real users and the attackers would and could re-broadcast their unconfirmed transactions to the network. IIRC, by default, bitcoin core will do so every half an hour.
So, I am not sure if this "remove from mempool" helps to counter or reduce the problem.

I'm not comparing it, I'm asking.... :-p

I'm not so much into the concept of sidechains, but wouldn't it possible that some players who send out lots of small transactions confirm them in a sidechain which then forwards one transaction in the blockchain? Wouldn't this make things easier?

yes, i always ask why ... they don't use the real 1Mb size : https://bt.irlbtc.com/view/1096975.msg11687498#msg11687498

 

Nope, Antpool is also not using the full size.
Some weeks ago, I read an article about this block limit the miners(pools) set them self, but I obviously haven't bookmarked it.
I have never heard of Discus Fish, but they seem to make the blocks as full as possible.

the "included within 6 blocks" is no indicator. I see this since 9 hours at blockchain.info

just hoard it then.

This could give legitimate users a choice to re-broadcast the transaction with higher fee or wait until another time. For attackers it does not matter since they are going to flood the network anyway

Again there is no universal solution against attacks. Chinese built a great wall, but the general that guarded the gate opened it for the enemy because of a woman

Reject high-frequency transactions from unknown source might be a better solution, but that means each node will be updated with a list for trusted high-frequency transaction entities, thus some kind of management scheme will be setup. Then there will be politics around who should or should not get into that list, will get complicated

Pure market driven solutions will not work due to some people who control large amount of resources will be able to attack for months while majority of users will not be able to compete with a higher fee. 1% of people having 99% of the resource is a fact