Pages:
Author

Topic: Time to download the block chain (Read 3084 times)

legendary
Activity: 1904
Merit: 1002
October 30, 2011, 07:24:16 PM
#27
There's no reason each release couldn't come with a torrent file with a recent snapshot of the blockchain.
hero member
Activity: 812
Merit: 1000
October 30, 2011, 07:19:19 PM
#26
Download speed of the blockchain is ridicolous and make no sense.

It's a P2P network where there is a 900MB file and pretty much everyone is a seeder. On bittorrent that mean FULL STEAM AHEAD! DOWNLOAD AT FULL SPEED!
It should take like 5 or 10 minutes.

Instead here it take 12+hours.......

a bittorrent download is done completely out of order and the client just continually fills in the blanks.

a blockchain download is done in order and only after block 123 is downloaded will it download block 124*

in addition, file compression would be far more efficient when compressing the entire chain as one big file (for possible deployment on bittorrent), whereas because bitcoin downloads each block separately, the compression over the whole chain is less efficient.

*could be wrong, please get a second opinion.
legendary
Activity: 1862
Merit: 1011
Reverse engineer from time to time
October 30, 2011, 07:02:52 PM
#25
No i downloaded an archive and waited only 2-3 minutes.

http://bitcoin.bluematt.me/bitcoin-nightly/blockchain-nightly/

I am not going to download the block chain from some random url.
Whatever...BlueMatt is a developer working on Bitcoin and this can be verified on freenode @ #bitcoin-dev
newbie
Activity: 53
Merit: 0
October 28, 2011, 11:23:12 AM
#24
Of course they don't because UPnP only works from LAN side.  I don't use it but I find the hysteria over UPnp to be amusing.  For it to present a risk you would already need malicous software inside the network.

You should read the FAQ on the umap pentest tool: http://www.toor.do/
legendary
Activity: 1512
Merit: 1049
Death to enemies!
October 27, 2011, 09:18:29 PM
#23
There is a lot of misguided sysadmins who think UPnP is malicius and vector of vulnerability. In reality UPnP can only be used to help attacker, when he got the payload already in the LAN side and the malware is run. It's game-over scenario anyway.

One more thing - with properly written UPnP software, the opened port rule is closed once the application using it is closed. With permanently forwarded port the port is forwarded at all times.
donator
Activity: 1218
Merit: 1079
Gerald Davis
October 27, 2011, 09:14:37 PM
#22

also, not all routers allow the user to just disable UPnP on wan or lan individually.

Of course they don't because UPnP only works from LAN side.  I don't use it but I find the hysteria over UPnp to be amusing.  For it to present a risk you would already need malicous software inside the network.  If that is true you likely are fucked UPnP or not.  The very people who are the most likely to allow that to happen (look an exe from my friend... click) have no chance to configuring a router manually anyways.  So without UPnp they are likely going to turn off all firewalls, put their PC on DMZ, or plug directly into cable modem which are far worse than enabling UPnP.   Also it isn't like lack of UPnP makes the network invulnerable to attack.

  

legendary
Activity: 1512
Merit: 1049
Death to enemies!
October 27, 2011, 09:14:11 PM
#21
DO NOT USE UPnP!!!!!!

in some routers, they can be completely reconfigured by sending only 1 packet. this can be solved by simply turning off UPnP. id also advise turning off all other remote and automatic configuration systems, they are only security holes.
In my systems I configure ports manually, and no UPnP. For most users who are clueless about networking, the router with UPnP is still more secure than going directly to internet. It's like going naked on street when gay pride is marching nearby.
Quote
If you have malicous software on the inside of your firewall/router then you have larger problems then UPnP.
Yes, the golden rule. It's either completely clean and safe, or you are owned. And Yes, most routers allow changing UPnP rules only from LAN side. Still there are routers with badly written firmware or running on default passwords and with WAN management acess enabled.
sr. member
Activity: 350
Merit: 251
October 27, 2011, 09:11:33 PM
#20
DO NOT USE UPnP!!!!!!

in some routers, they can be completely reconfigured by sending only 1 packet. this can be solved by simply turning off UPnP. id also advise turning off all other remote and automatic configuration systems, they are only security holes.

Um UPnP can only be enabled by an application on the LAN side.  If you have malicous software on the inside of your firewall/router then you have larger problems then UPnP.

Also is the number of packets suppose to be scary?  If it took 500 packets to enable UPnP would that somehow be safer than a single packet?

if you need UPnP then you probably have no idea what lan and wan is, and probably no idea what UPnP even is. if you know all of them terms then you should have no trouble at all forwarding the ports manually, it literally takes 2 minutes if you are familiar with your routers UI. also, it would not really matter the amount of packets required, the exploit still exists and is just as dangerous.

also, not all routers allow the user to just disable UPnP on wan or lan individually.
legendary
Activity: 1386
Merit: 1097
October 27, 2011, 09:11:06 PM
#19
Yesterday I downloaded blockchain on 10Gbit line, few Xeon CPUs and SSD raid. It took only one hour to complete :-).

I followed iostat, iftop and htop and it definitely looks that the biggest issue is waiting on data from network. Sometimes it processed 500 blocks per second, but then suddently dropped connection to peer and it took long time (minutes?) to recover and continue downloading.
donator
Activity: 1218
Merit: 1079
Gerald Davis
October 27, 2011, 09:03:54 PM
#18
DO NOT USE UPnP!!!!!!

in some routers, they can be completely reconfigured by sending only 1 packet. this can be solved by simply turning off UPnP. id also advise turning off all other remote and automatic configuration systems, they are only security holes.

Um UPnP can only be enabled by an application on the LAN side.  If you have malicous software on the inside of your firewall/router then you have larger problems then UPnP.

Also is the number of packets suppose to be scary?  If it took 500 packets to enable UPnP would that somehow be safer than a single packet?
sr. member
Activity: 350
Merit: 251
October 27, 2011, 09:00:45 PM
#17
DO NOT USE UPnP!!!!!!

in some routers, they can be completely reconfigured by sending only 1 packet. this can be solved by simply turning off UPnP. id also advise turning off all other remote and automatic configuration systems, they are only security holes.
legendary
Activity: 1512
Merit: 1049
Death to enemies!
October 27, 2011, 08:47:35 PM
#16
(yes, my internet connection is fine, and I have 8 connections).
No, Your connection is not fine. Open the necessary ports in router or enable UPnP. You will most likely get more speed. Also the power of computer and random acess times of harddrive is also important. I taken hour and a half to download blockchain on Core2Duo running on WD Raptor, but the same blocks taked whole day to download on aging Pentium4 running with old Maxtor drive.

In future there might be good idea to release the Bitcoin software with two installers - the standart and the Mega installer, that installs the blockchain files in Bitcoin directory.
legendary
Activity: 1414
Merit: 1000
HODL OR DIE
October 27, 2011, 04:16:11 PM
#15
if you only get 8 connections you need to open some ports
legendary
Activity: 1246
Merit: 1016
Strength in numbers
October 27, 2011, 03:47:11 PM
#14
Download speed of the blockchain is ridicolous and make no sense.

It's a P2P network where there is a 900MB file and pretty much everyone is a seeder. On bittorrent that mean FULL STEAM AHEAD! DOWNLOAD AT FULL SPEED!
It should take like 5 or 10 minutes.

Instead here it take 12+hours.......
I think the client limits its connections to a few tens of peers. Given that these are random peers, their upload speeds are going to be horrible. (This is speculation though, I may be wrong about how this works.)

This seems right, but also there is a huge difference on a netbook compared to a desktop. I think this is because it validates as blocks come in and machine speed matters.
donator
Activity: 1218
Merit: 1079
Gerald Davis
October 27, 2011, 03:44:52 PM
#13
No i downloaded an archive and waited only 2-3 minutes.

http://bitcoin.bluematt.me/bitcoin-nightly/blockchain-nightly/

I am not going to download the block chain from some random url.


Your bitcoin client will check all the blocks and if they are nonvalid, it won't accept them. There is no good way to fake the block chain, the whole "hardcode-first-block-hash-and-then-include-hash-of-previous-block-in-each-next-block" thing is what makes bitcoin so secure. So unless there is some 0-day bug in the bitcoin client I wouldn't worry too much about downloading block chain from unknown source.

Erhm... When you download the database as a bunch of files and start the client it is NOT validating it. This would mean that the client would have to validate all transactions whenever it is started.

Well it depends.  The db contains a record of last block verified.  The person uploading the db could modify it so that is block 0 and the client would rescan entire block chain.  Alternatively use the -rescan command line option to force a complete verification of downloaded block chain.  Hopefully in future client will have a "verify blockchain" option/menu/button.

I do agree this could be improved. 
Why not include 70% of the blockchain with the binary?
Why not give client an option to download entire block chain up to a week prior from a repository and then auto-rescan?
legendary
Activity: 1148
Merit: 1008
If you want to walk on water, get out of the boat
October 27, 2011, 02:27:56 PM
#12
Maybe you are right

hero member
Activity: 714
Merit: 504
^SEM img of Si wafer edge, scanned 2012-3-12.
October 27, 2011, 02:25:14 PM
#11
Download speed of the blockchain is ridicolous and make no sense.

It's a P2P network where there is a 900MB file and pretty much everyone is a seeder. On bittorrent that mean FULL STEAM AHEAD! DOWNLOAD AT FULL SPEED!
It should take like 5 or 10 minutes.

Instead here it take 12+hours.......
I think the client limits its connections to a few tens of peers. Given that these are random peers, their upload speeds are going to be horrible. (This is speculation though, I may be wrong about how this works.)
legendary
Activity: 1148
Merit: 1008
If you want to walk on water, get out of the boat
October 27, 2011, 02:23:35 PM
#10
Download speed of the blockchain is ridicolous and make no sense.

It's a P2P network where there is a 900MB file and pretty much everyone is a seeder. On bittorrent that mean FULL STEAM AHEAD! DOWNLOAD AT FULL SPEED!
It should take like 5 or 10 minutes.

Instead here it take 12+hours.......
hero member
Activity: 868
Merit: 1008
October 27, 2011, 01:57:26 PM
#9
Downloading from the p2p network shouldn't be that slow compared with downloading from a website...verification will take time, but that should be done regardless of how the chain is first downloaded.  Maybe the client should try to connect to more peers when it's catching up, dropping slow peers and trying to latch on to faster ones.
Jan
legendary
Activity: 1043
Merit: 1002
October 27, 2011, 01:45:33 PM
#8
No i downloaded an archive and waited only 2-3 minutes.

http://bitcoin.bluematt.me/bitcoin-nightly/blockchain-nightly/

I am not going to download the block chain from some random url.


Your bitcoin client will check all the blocks and if they are nonvalid, it won't accept them. There is no good way to fake the block chain, the whole "hardcode-first-block-hash-and-then-include-hash-of-previous-block-in-each-next-block" thing is what makes bitcoin so secure. So unless there is some 0-day bug in the bitcoin client I wouldn't worry too much about downloading block chain from unknown source.

Erhm... When you download the database as a bunch of files and start the client it is NOT validating it. This would mean that the client would have to validate all transactions whenever it is started.
Pages:
Jump to: