mas.
s62
lecture 5
synchronization
2018-02-21
Tadge Dryja
1
the Bitcoin network
so far we've talked about:
signatures
mining and blocks
transactions and scripts
... now to put it all together
2
recap: signatures
public / private keys
private key can sign() a message
can verify(public key, message, sig)
useful for proving identity,
ownership. Better than paper
signatures!
3
recap: mining and blocks
change a nonce, hash a bunch of
times, get a low output. Proves work
Include the previous data as part of
your input, and you make a chain of
work -- a blockchain
4
recap: txs and scripts
Transactions have inputs and outputs
inputs outputs
txid:index (36B) script (25B)
signature (100B) amount (8B)
txid:index script (pubkey)
signature amount
5
recap: txs and scripts
inputs point to old outputs and have
signatures
outputs have scripts and coin amounts
txid:index
inputs (36B) script (25B)
outputs
signature (100B) amount (8B)
txid:index script (pubkey)
signature amount
6
tx mining process
users make txs, sign, broadcast
someone takes all the txs, puts them
in a block, and does work
those txs are now "confirmed", and
the next block can be built
7
tx mining: header
the block header is the message which
must satisfy the proof of work
Headers have a hash of the txs in the
block
Really it's the headers that make a
chain, not the blocks. Headerchain.
8
tx mining: header
headers are 80 bytes; similar to
pset02 blocks
Main components are
prev hash, merkle root, nonce
9
merkle root recap
Hash in a binary
tree
Same level of hash
h,h
commitment as
h(0,1,2,3) hash
0,1
hash
2,3
txid txid txid txid
0 1 2 3
10
header fields
version 4B indicates block
prev hash 32B version
merkle root 32B Was used for fork
time 4B signalling;
diff 4B future use
nonce 4B unclear
11
header fields
version 4B hash of previous
prev hash 32B block
merkle root 32B
time 4B
diff 4B
nonce 4B
12
header fields
version 4B hash of all
prev hash 32B transactions in
merkle root 32B the block
time 4B
diff 4B
nonce 4B
13
header fields
version 4B unix time
prev hash 32B (seconds since
merkle root 32B 1970) of claimed
time 4B block creation
diff 4B
nonce 4B (can be before
previous block's
time!)
14
header fields
version 4B PoW target in a
prev hash 32B weird floating
merkle root 32B point format
time 4B
diff 4B pretty much
nonce 4B useless as can be
computed anyway
15
header fields
version 4B nonce - anything
prev hash 32B goes here
merkle root 32B
time 4B but there's a
diff 4B problem...
nonce 4B
16
header fields
version 4B nonce - anything
prev hash 32B goes here
merkle root 32B
time 4B but there's a
diff 4B problem...
nonce 4B
too small!
17
header fields
version 4B 232 possible
prev hash 32B nonces
merkle root 32B
time 4B But current
diff 4B blocks need 270
nonce 4B work!
18
header fields
version 4B adjust time
prev hash 32B
merkle root 32B modify merkle
time 4B root
diff 4B
nonce 4B
19
tx order in block
tx0 is the coinbase tx:
generates new coins, and takes fees
from all other txs in block
all other txs can be in any order,
but can only spend outputs from
previous txs
20
tx order in block
if txB spends an output of txA, then
txA must come first in block ordering
this ensures linear verification of
transactions can proceed
21
intermission
256 second break
prove work by moving body mass
against force of gravity
work = f*d = m*g*h
22
sync process
I just downloaded bitcoin!
What's been going on for the last 9
years?
23
sync process
Download binary / compile code
Verify GPG signatures somehow...
Hardcoded DNS seeds to find peers
connect, ask for headers
download & verify 500K headers
24
sync process
Get the header chain first - quick
takes under a minute with good
connections
verify all the work before any
signatures
25
sync IBD
After headers, Initial Block Download
(IBD)
Request blocks from peers, match tx
list to merkle root in header,
process each tx in order
26
sync IBD
Delete all input txos
TX
inputs outputs
UTXO DB
27
sync IBD
Delete all input txos
Add output txos TX
inputs outputs
UTXO DB
28
sync IBD
Do this ~300M times
Downloads 170GB
End result:
UTXO DB
55M txos, ~3.2GB
29
pruning
By default, store all 500K blocks
Can serve to others who need to IBD
But can "prune" / delete blocks after
IBD with no loss of security
Downside?
30
pruning
By default, store all 500M blocks
Can serve to others who need to IBD
But can "prune" / delete blocks after
IBD with no loss of security
Downside? Not everyone can prune
31
blockchain data
What does it store?
banlist.dat
chainstate
peers.dat
wallet.dat
bitcoin.conf
blocks
database
debug.log
mempool.dat 32
blockchain data
What does it store?
banlist.dat 1.8K bad nodes
chainstate
peers.dat 4.0M good nodes
wallet.dat 1.4M my precious
bitcoin.conf 144 config file
blocks
database
debug.log 11M log file, rotates
mempool.dat 20M more like diskpool 33
blockchain data
What does it store?
banlist.dat 1.8K
chainstate 3.0G utxo set
peers.dat 4.0M
wallet.dat 1.4M
bitcoin.conf 144
blocks 183G all the
blocks
database 80K ? nothing?
debug.log 11M 34
blockchain as database
186GB, but a really crummy database
remember tx 9e95c3c3c96f57527cdc649550bf8e92892f7651f718d846033798aee333b0c3
from back in 2014?
35
blockchain as database
186GB, but a really crummy database
remember tx 9e95c3c3c96f57527cdc649550bf8e92892f7651f718d846033798aee333b0c3
from back in 2014?
No. It's somewhere in the blocks
folder but I don't know where.
It's not in chainstate 36
blockchain as database
how about output
02b1500a0f3b059819dd923f1c78bacc0a3de303fc51836ce7f46a3206b29ba7:0
it's an op_return output, can you tell
me what the data is?
37
blockchain as database
how about output
02b1500a0f3b059819dd923f1c78bacc0a3de303fc51836ce7f46a3206b29ba7:0
it's an op_return output, can you tell
me what the data is?
Nope! op_return outputs don't get
stored in the chainstate.
38
blockchain as database
Hey I have a pubkey with hash
1d493f9536c692d096536ba9d1c081feabd7ccf3
how many coins do I have? How many
outputs?
39
blockchain as database
Hey I have a pubkey with hash
1d493f9536c692d096536ba9d1c081feabd7ccf3
how many coins do I have? How many
outputs?
No idea! Gotta search through all of
chainstate. Doesn't index based on
PkScript, only txid:index 40
blockchain as database
how many coins does output
7434e09a302eaa4e2e0826aea08c2cca282a8bfc606cb680aa1f3f331a7e4f69:1
have?
41
blockchain as database
how many coins does output
7434e09a302eaa4e2e0826aea08c2cca282a8bfc606cb680aa1f3f331a7e4f69:1
have?
Lots! 239.99913132. It's in the utxo
set because it hasn't been spent yet.
Can quickly find based on txid:index
42
blockchains are bad databases
Only keeps track of utxos, which is
hard enough
Can add further indexes, but they take
lots of space. Most common is
"address index" so people can ask if
they have any money.
43
blockchains are bad databases
DB queries not given to network peers
Network peers are scary, ban them if
they act funny
Provide headers, blocks, txs, other
nodes IPs
44
bad DB but good consensus
Everyone's got the same utxo set
Even though they all really want more
utxos. Or to break the system. It
seems to work.
45
pset02 update
Bunch of blocks mined
Pls reduce server queries; an 18.
address is doing 5+ TCP connections
per second. Also GCE? 35.
Could use blocks here to start a
coin...
46
MIT OpenCourseWare
https://ocw.mit.edu/
MAS.S62 Cryptocurrency Engineering and Design
Spring 2018
For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.
