From 370,000 Full Bitcoin Nodes to 6,000: What happened?

[coins101]

....
The main problem is that there is no incentive to be a node and there is not enough (is for now) people who would voluntarily run nodes.

There is an incentives program being planned, but it is too early yet to say if it will work technically as paying nodes is difficult without Sybil attackers spoofing the network to steal the incentives.

It is also not clear if the proposals will be generally accepted.

[qwk]

It doesn't matter if it was counting QT users or not, because it is a fact that the number of nodes has definitely fallen over time and it might start reaching critical levels at some point.

There's not really something like a critical level of nodes.
Well, obviously 0 is too few. Anything above a couple handful is sufficient to keep the blockchain running without too much of a risk of losing the data at some point.
For practical purposes of course, we need a good distribution of nodes to facilitate lightweight use of the blockchain.

[Gleb Gamow]

WOW! With their 102 Nodes and decreasing, I now see why Gavin and Hearn are having a hard time tryin' to convince the Chinese...  

[Sythyn]

blockchain is becoming bigger and more encumbant also service rate of bitcoin is dropping because of the price

[disclosure]

The early December 2013 network snapshots, i.e. with over 100k nodes, are not valid as the crawler at that time took several hours just to complete one full network snapshot and it includes all nodes from addr responses which can be faked or likely stale. Those snapshots were linked to from getaddr.bitnodes.io/SNAPSHOT_NUMBER/ which have been removed since.

With the new crawler released in late Dec 2013, the crawl time was brought down significantly down to sub 5 minutes with considerably good churn rate. I use the churn rate to measure how good a network snapshot is. A value of 0 implies the network snapshot was taken instantly but of course that's not possible. At the moment, we are seeing a typical churn rate of 30+: https://getaddr.bitnodes.io/nodes/

[Lauda]

....
The main problem is that there is no incentive to be a node and there is not enough (is for now) people who would voluntarily run nodes.

There is an incentives program being planned, but it is too early yet to say if it will work technically as paying nodes is difficult without Sybil attackers spoofing the network to steal the incentives.

It is also not clear if the proposals will be generally accepted.

I think that something is definitely going to be needed. However we should look into a lot of options, and try some out. Maybe even crowdfund a few hundreds nodes if we really need them?

There's not really something like a critical level of nodes.
Well, obviously 0 is too few. Anything above a couple handful is sufficient to keep the blockchain running without too much of a risk of losing the data at some point.
For practical purposes of course, we need a good distribution of nodes to facilitate lightweight use of the blockchain.

Considering the amount of users that Bitcoin has today, do you really think that it would work with 1 single relay node? I don't even want to talk about the bandwidth issues.
What I'm more concerned is the security. With only a single node wouldn't the network be more susceptible to sybil attacks?
Wouldn't it be more easy to DDoS that single node and take the network down for hours (i.e. no relaying of blocks)?

[Sitarow]

blockchain is becoming bigger and more encumbant also service rate of bitcoin is dropping because of the price

The reality is that we no longer mine Bitcoin using computers CPU's and GPU's. Hence the reduced chances of people installing bitcoin client and becoming a full node even tho they end up using the system to mine on a pool.

[coins101]

The early December 2013 network snapshots, i.e. with over 100k nodes, are not valid as the crawler at that time took several hours just to complete one full network snapshot and it includes all nodes from addr responses which can be faked or likely stale. Those snapshots were linked to from getaddr.bitnodes.io/SNAPSHOT_NUMBER/ which have been removed since.

With the new crawler released in late Dec 2013, the crawl time was brought down significantly down to sub 5 minutes with considerably good churn rate. I use the churn rate to measure how good a network snapshot is. A value of 0 implies the network snapshot was taken instantly but of course that's not possible. At the moment, we are seeing a typical churn rate of 30+: https://getaddr.bitnodes.io/nodes/

That's a definitive answer.

Thanks for helping resolve the issue.

[Sitarow]

The early December 2013 network snapshots, i.e. with over 100k nodes, are not valid as the crawler at that time took several hours just to complete one full network snapshot and it includes all nodes from addr responses which can be faked or likely stale. Those snapshots were linked to from getaddr.bitnodes.io/SNAPSHOT_NUMBER/ which have been removed since.

With the new crawler released in late Dec 2013, the crawl time was brought down significantly down to sub 5 minutes with considerably good churn rate. I use the churn rate to measure how good a network snapshot is. A value of 0 implies the network snapshot was taken instantly but of course that's not possible. At the moment, we are seeing a typical churn rate of 30+: https://getaddr.bitnodes.io/nodes/

That's a definitive answer.

Thanks for helping resolve the issue.

The above explanation would have provide noticeable results quickly, however it does not account the observable drop in nodes over time.

https://bitinfocharts.com/comparison/nodes-btc-ltc.html#log

You can see that there are serious technology implications and mining adoption pertaining to full BTC nodes and the migration from GPU mining to ASIC.

P.S. As to your question "Q3. Is is possible to ever see that number again?"

Introducing incentive by granting % of transaction fee's to the node that relays may help.

[qwk]

There's not really something like a critical level of nodes.
Well, obviously 0 is too few. Anything above a couple handful is sufficient to keep the blockchain running without too much of a risk of losing the data at some point.
For practical purposes of course, we need a good distribution of nodes to facilitate lightweight use of the blockchain.

Considering the amount of users that Bitcoin has today, do you really think that it would work with 1 single relay node? I don't even want to talk about the bandwidth issues.

No, it would not "work" in the sense that it'd be usable.
But it could still accept transactions, let miners verify them and append new blocks to the chain, technically.

All it takes for bitcoin to survive though, is really just a single copy of the blockchain, not even a running node.

What I'm more concerned is the security. With only a single node wouldn't the network be more susceptible to sybil attacks?
Wouldn't it be more easy to DDoS that single node and take the network down for hours (i.e. no relaying of blocks)?

Security is mostly dependent upon the miners, not really the nodes. A hypothetical network of just a single node with hundreds of miners would still be "secure". Sybil attacks would be easy, sure. An attacker could "forge" his own chain by only relaying his own blocks, thereby creating his own network. It'd be hard to tell which one is the "right" one. If you find yourself on the wrong side, you might be in trouble

To sum it up, a network with a decent number of nodes is really desirable.

[Amitabh S]

Do bitcoinj (SPV) nodes count in this number?

[EternalWingsofGod]

Earlier running a full node was the only way to have a wallet. Now, there are many lightwight clients and online wallets to give people the ease of use. Naturally people have stopped running a full node. Now-a-days full nodes are mainly run by those who need it, i.e. business to verify their Tx and miners. There are few volunteers who run full node though.

That is part of the reason, other factors invlove is that the upload download speed people have which impacts the time it takes to download the full client and people tend to install the client from scratch instead of using a snapshot which takes time, besides the size of the Blockchain  light weight clients are seen as more attractive to have because of size and time constraints.

[Kprawn]

The early December 2013 network snapshots, i.e. with over 100k nodes, are not valid as the crawler at that time took several hours just to complete one full network snapshot and it includes all nodes from addr responses which can be faked or likely stale. Those snapshots were linked to from getaddr.bitnodes.io/SNAPSHOT_NUMBER/ which have been removed since.

With the new crawler released in late Dec 2013, the crawl time was brought down significantly down to sub 5 minutes with considerably good churn rate. I use the churn rate to measure how good a network snapshot is. A value of 0 implies the network snapshot was taken instantly but of course that's not possible. At the moment, we are seeing a typical churn rate of 30+: https://getaddr.bitnodes.io/nodes/

That's a definitive answer.

Thanks for helping resolve the issue.

This, and what Mike Hearn said about the crawler being more effective now. The crawler software was improved to only count nodes with 100% access and fully configured. It now looks at different parameters before it counts it as a full node. {If it's poorly configured and does not give full access, it will not be counted}

So there might be people out there, who have poorly configured nodes, and they do not even know it.

This should be addressed by the Bitcoin Core developers {Setting of default parameters, to automatically configure nodes to run at full capacity or minimum standards to be counted as a full node}

People just install the software and run the default options. 

[sickpig]

AFAIK the way to measure the # of nodes has been changed along the way. The crawler is actually counting nodes accepting incoming connections on port 8333. Earlier version takes into account also all of them.

That said the increased burden in running a full node had definitely contributed to the trend.

Unfortunately we don't know how much effect the increase size of the  blocksize has had, due to the change  in the way we observed the nodes number.

[coins101]

AFAIK the way to measure the # of nodes has been changed along the way. The crawler is actually counting nodes accepting incoming connections on port 8333. Earlier version takes into account also all of them.

That said the increased burden in running a full node had definitely contributed to the trend.

Unfortunately we don't know how much effect the increase size of the  blocksize has had, due to the change  in the way we observed the nodes number.

I'm hearing this more and more.

It's also something that needs to be factored into the block size debate, as well as any incentives program.

[Denker]

Guys please help me out. What if 21 inc. starts to bring out it's embedded mining chips for smartphones, routers, and several other devices? To mine some dust of coins these devices have to run a full node client right? I mean this COULD mean an explosion of network nodes or how will they be able to mine otherwise?

[Lauda]

Guys please help me out. What if 21 inc. starts to bring out it's embedded mining chips for smartphones, routers, and several other devices? To mine some dust of coins these devices have to run a full node client right? I mean this COULD mean an explosion of network nodes or how will they be able to mine otherwise?

This isn't correct. First of, I don't see the benefit in this. Anyone who mines with any of these devices would be doing so at a loss. Essentially the people that would mine like that are the ones that truly support the network.
You don't understand this correctly. There are 2 cases:
If you're solo mining, then you need the entire blockchain. If you don't have the last block yet, the network will most likely reject your mined blocks since you're already behind the chain.
If you're mining in a pool, then you don't need to download the blockchain. You just point your miner at a pool and start mining.

I hardly doubt that there is going to be enough storage in those devices to run full nodes as the blockchain is already at ~40GB.

No, it would not "work" in the sense that it'd be usable.
But it could still accept transactions, let miners verify them and append new blocks to the chain, technically.

All it takes for bitcoin to survive though, is really just a single copy of the blockchain, not even a running node.
To sum it up, a network with a decent number of nodes is really desirable.

Yeah, we weren't really going in the same direction with this. I was trying to point out that it wouldn't be usable with a very low number of nodes. However, I do agree with you as our conclusion seems to be the same.

[qwk]

Guys please help me out. What if 21 inc. starts to bring out it's embedded mining chips for smartphones, routers, and several other devices? To mine some dust of coins these devices have to run a full node client right? I mean this COULD mean an explosion of network nodes or how will they be able to mine otherwise?

As LaudaM already pointed out, those devices will not run a full node.
They'll be workers in a mining pool.
What that means is that somewhere at 21 inc they'll have one full node and all those little machines will share their computing power with that single node. Well, most probably more than one, but you get the idea.

[Denker]

Guys please help me out. What if 21 inc. starts to bring out it's embedded mining chips for smartphones, routers, and several other devices? To mine some dust of coins these devices have to run a full node client right? I mean this COULD mean an explosion of network nodes or how will they be able to mine otherwise?

As LaudaM already pointed out, those devices will not run a full node.
They'll be workers in a mining pool.
What that means is that somewhere at 21 inc they'll have one full node and all those little machines will share their computing power with that single node. Well, most probably more than one, but you get the idea.

Thanks for clearing things up. I wasn't sure how this will work that's why I asked.

[ranochigo]

It seems to me that there needs to be a distinction between "running a full node" and running Bitcoin Core.

There may have been 370k people running Bitcoin Core, but how much benefit were they providing to the network? Running Bitcoin Core long enough to sync the block chain and send some bitcoins or view your balance provides very little benefit, especially if you haven't opened port 8333.

If you have over 8 connections one can assume that port is open?

By default, Bitcoin Core only allows 8 outgoing connections. To attain more than that, port 8333 must be open. However, you can still modify net.cpp line 57 and change the max connections. The best way to check is by going to https://getaddr.bitnodes.io/ and check your connection by using the join the network.