Difference between revisions of "P2Pool"
(fixed phoenix miner link)
m (Spelling fix.)
|Line 48:||Line 48:|
===== cgminer =====
===== cgminer =====
A workaround was needed to get long polling to work, but it has
A workaround was needed to get long polling to work, but it has fixed long polling delay. Works perfectly (with the latest p2pool).
===== ufasoft =====
===== ufasoft =====
Revision as of 08:53, 9 August 2011
These nodes work on a chain of shares similar to Bitcoin's blockchain. Each node works on a block that includes payouts to the previous shares' owners and the node itself. There is no central point of failure, making it DoS resistant.
Miners are configured to connect to a P2Pool node that can be run locally, alongside the miner.
- 1 Overview
- 2 Joining the pool
- 3 Protocol description
- 4 History
- 5 See Also
- 6 External Links
- 7 References
P2Pool shares form a "sharechain" with each share referencing the previous share's hash. Each share contains a standard Bitcoin block header, some P2Pool-specific data that is used to compute the generation transaction (total subsidy, payout script of this share, a nonce, the previous share's hash, and the current target for shares), and a merkle branch linking that generation transaction to the block header's merkle hash.
The chain continuously regulates its target to keep generation around one share every five seconds.
Unlike Bitcoin, nodes do not know the entire chain - instead they only hold the last 17280 shares (the last day's worth). In order to prevent an attacker from working on a chain in secret and then releasing it, overriding the existing chain, chains are judged by how much work they have since a point in the past. To ascertain that the work has been done since that point, nodes look at the Bitcoin blocks that the shares reference, establishing a provable timestamp. (If a share points to a block, it was definitely made after that block was made.)
Each share contains a generation transaction that pays out to the previous n shares, where n is the number of shares whose total work is equal to 3 times the average work required to solve a block, or 17280, whichever is smaller. Payouts are weighted based on the amount of work each share took to solve - proportional to the p2pool difficulty at that time.
The block reward (currently 50BTC) and the transaction fees are combined and apportioned according to these rules:
A subsidy of 0.5% is sent to the node that solved the block in order to discourage not sharing solutions that qualify as a block. (An evil miner could withhold the block, thereby preventing anybody from getting paid. He can NOT redirect the payout to himself.) In addition, a subsidy of 0.5% is kept as a fee. The remaining 99% is distributed evenly to miners based on work done recently.
In the event that a share qualifies as a block, this generation transaction is exposed to the Bitcoin network and takes effect, transferring each node its payout.
Joining the pool
Follow these steps to join the pool:
- Run bitcoin with the RPC interface enabled: put rpcuser=USER, rpcpassword=PASS, and server=1 in bitcoin.conf
- Download p2pool:
- Windows py2exe: http://u.forre.st/u/rewbmxxo/p2pool_win32_fbedaf9.zip
- git: git clone git://github.com/forrestv/p2pool.git
- Run p2pool: (See below for additional options.)
- Windows py2exe: run_p2pool.exe USER PASS
- git: python run_p2pool.py USER PASS
- Run a miner daemon with long polling connecting to 127.0.0.1 (or the IP of the host running p2pool if you're mining remotely) on port 9332 with any username and password
- poclbm.exe http://a:firstname.lastname@example.org:9332
Dependencies if using version from git:
- python-argparse for Python 2.6 and lower
- Twisted (Ubuntu package python-twisted)
This is all for the latest p2pool version, it includes several new workarounds. With all miners, using a high FPS target (100?) or a low aggression, and waiting for p2pool to completely start (line shows > 35000 shares) helps a lot with reducing stales.
A workaround was needed to get long polling to work, but it has now fixed the long polling delay. Works perfectly (with the latest p2pool).
By looking at the source but not actually running it (couldn't get it to compile) should work perfectly too.
Needs patch to reduce stales
Has a bug where it waits half a second after receiving a long polling response. There is a 20% chance that a p2pool share will be generated in that window, resulting in all hashing until the next getwork being useless.
A fork of DiabloMiner on GitHub that fixes this: https://github.com/forrestv/DiabloMiner Make sure to get the fast_long_polling branch!
git clone git://github.com/forrestv/DiabloMiner.git cd DiabloMiner git checkout fast_long_polling
Built with JAR and Windows binary: http://u.forre.st/u/xsvenrwa/DiabloMiner_2e73cc.zip
Has a similar bug, but it waits a full second.
git clone git://github.com/forrestv/poclbm.git cd poclbm git checkout fast_long_polling
No binary as of yet.
Has a similar bug (waiting a full second).
GitHub fork: https://github.com/forrestv/phoenix-miner
Windows binary: http://u.forre.st/u/ycvzejfm/phoenix-miner-fba48d2-5.zip
puddinpop's RPC Miners
Does not have long polling so will not work at all.
usage: run_p2pool.py [-h] [--version] [--testnet] [--debug] [-a ADDRESS] [--p2pool-port PORT] [-n ADDR[:PORT]] [-l] [-w PORT] [--bitcoind-address BITCOIND_ADDRESS] [--bitcoind-rpc-port BITCOIND_RPC_PORT] [--bitcoind-p2p-port BITCOIND_P2P_PORT] BITCOIND_RPC_USERNAME BITCOIND_RPC_PASSWORD p2pool (version a07e25c18d122565793536945c8cc65049eb1cce) optional arguments: -h, --help show this help message and exit --version show program's version number and exit --testnet use the testnet --debug debugging mode -a ADDRESS, --address ADDRESS generate to this address (defaults to requesting one from bitcoind) -l, --low-bandwidth trade lower bandwidth usage for higher latency (reduced efficiency) p2pool interface: --p2pool-port PORT use TCP port PORT to listen for connections (default: 9333 normally, 19333 for testnet) (forward this port from your router!) -n ADDR[:PORT], --p2pool-node ADDR[:PORT] connect to existing p2pool node at ADDR listening on TCP port PORT (defaults to 9333 normally, 19333 for testnet), in addition to builtin addresses worker interface: -w PORT, --worker-port PORT listen on PORT for RPC connections from miners asking for work and providing responses (default: 9332) bitcoind interface: --bitcoind-address BITCOIND_ADDRESS connect to a bitcoind at this address (default: 127.0.0.1) --bitcoind-rpc-port BITCOIND_RPC_PORT connect to a bitcoind at this port over the RPC interface - used to get the current highest block via getwork (default: 8332) --bitcoind-p2p-port BITCOIND_P2P_PORT connect to a bitcoind at this port over the p2p interface - used to submit blocks and get the pubkey to generate to via an IP transaction (default: 8333 normally. 18333 for testnet) BITCOIND_RPC_USERNAME bitcoind RPC interface username BITCOIND_RPC_PASSWORD bitcoind RPC interface password
P2Pool's protocol mirrors Bitcoin's P2P protocol in many ways. It uses the same framing (prefix, command, length, checksum, payload) and similar commands:
- version - sent to establish a connection - contains (version, services, addr_to, addr_from, nonce, sub_version, mode, best_share_hash)
- setmode - sent to update the mode sent in the version message - contains (mode)
- ping - sent to keep connection alive - contains ()
- addrme - request that the receiving node send out an addr for the sending node - contains (port)
- addrs - broadcast list of nodes' addresses - contains (addrs)
- getaddrs - request that the receiving node send count addrs - contains (count)
- getshares - request that the receiving node send the shares referenced by hashes and parents of their parents, stopping at any share referenced by stops - contains (hashes, parents, stops)
- share0s - broadcast message of the hashes of shares - contains (hashes)
- share1as - broadcast message of the contents of shares that do not qualify as bitcoin blocks - contains (share1as)
- share1bs - broadcast message of the contents of shares that qualify as bitcoin blocks - contains (share1bs)
The mode is 0 for low-bandwidth mode, which sends the hash of a share first before sending the entire share, or 1 for low-latency (default) mode, which forwards the entire (~170 byte) share always. The two nodes on either side of a connection use the minimum of their two preferences for that connection.
This project was announced on June 17, 2011.
The software author's address for donations can be found in the signature section of his forum profile.