In our last post, we talked about what Proof of Work was and showed you how to code your own Proof of Work blockchain. The 2 most popular cryptocurrencies, Bitcoin and Ethereum are both based on Proof of Work.
But what are the downsides of Proof of Work? One of the major ones is electricity consumption. There is a race to set up bigger and bigger mining rigs to get the hardware power needed to mine more bitcoin. Check out the madness for yourself in this photo of a mining setup:
This costs an insane amount of electricity. Bitcoin mining alone consumes more energy than 159 countries! This is pretty irresponsible. However, from a technological perspective there are other downsides to Proof of Work. As more and more people participate in mining, the difficulty of the consensus algorithm needs to go up, creating a need for more hashing power. This means blocks and transactions take longer to get processed and get more expensive to mine. Proof of Work is a race to the bottom.
There are many thought leaders trying to find alternatives to Proof of Work. The most promising one so far is Proof of Stake. There are already production-ready blockchains based on Proof of Stake like Nxt and Neo. Ethereum is probably moving to Proof of Stake — their Casper project is already live on their test net.
So what the heck is Proof of Stake?
Instead of nodes competing with each other to solve hashes, in Proof of Stake, blocks are “minted” or “forged” (there is no “mining” so we don’t use that word in Proof of Stake) based on the amount of tokens each node is willing to put up as collateral. These nodes are called validators. We will be using the terms “nodes” and “validators” interchangeably in this tutorial. The tokens are specific to the blockchain. So in Ethereum, each node (validator) would put up Ether as collateral.
The more tokens each validator is willing to put up as collateral, the greater the chance they have to forge the next block and receive a reward. You can think of this as deposit interest. The more money you put in your savings account at your bank, the greater the monthly interest payment you receive.
Similarly, your probability of forging the next block increases the more tokens you put up as collateral. You are “staking” your tokens, which is why this consensus mechanism is called Proof of Stake.
What are the downsides to Proof of Stake?
You can probably already guess that a validator who has an excessively large amount of tokens staked will enjoy a disproportionately high probability of forging new blocks. However, this isn’t really different from what we see in Proof of Work. Bitcoin mining farms are getting so powerful that regular people haven’t been able to mine on their own laptops in years. Thus, many people argue that Proof of Stake is actually more democratized since anyone can at least participate on their own laptops without setting up a giant mining rig. They don’t need expensive hardware, just enough tokens to stake.
There are other downsides to Proof of Stake, both from a technological and economic standpoint. We won’t get into all of them here but here is a nice intro. In reality, Proof of Stake and Proof of Work both have their strengths and projects like Ethereum’s Casper blends features from both.
As always, the best way to understand how Proof of Stake works is to write your own code!
Let’s code a Proof of Stake blockchain!
We recommend checking out our networking tutorial before proceeding. It’s not compulsory but in some parts of the following tutorial we will be moving quickly, so it will help you to review it.
Our blockchain will implement the core concepts of Proof of Stake. However, because we need to be reasonable with length, the following production-level elements of a Proof of Stake blockchain will be left out.
- Full peer-to-peer implementation. Networking is simulated and the central blockchain state is held by a single Go TCP server. In this tutorial, the state is broadcast to each node from the single server.
- Wallet and balance tracking. We have not implemented a wallet in this code. Nodes are spun up in the network and the token amount is inputted in
stdin. So you can type in any amount you want. A full implementation would associate each node with a hash address and keep track of token balances in each.
- We’ll have a Go-based TCP server to which other nodes (validators) can connect.
- The latest blockchain state will get broadcast to each node periodically.
- Each node will propose new blocks.
- Based on the number of tokens staked by each node, one of the nodes will be randomly (weighted by the number of tokens staked) chosen as the winner, and their block will get added to the blockchain.
Setup and Imports
Before starting to write the code, we need to set an environment variable so our TCP server knows which port to use. Let’s create a
.env file in our working directory with one line in it:
Our Go program will read from this file and know to expose port 9000 so our nodes can connect to it.
Now let’s create a
main.go file in our working directory and start coding!
As usual, let’s make our package declaration and write up the imports we’ll need.
spewis a convenient package that pretty prints our blockchain to the terminal
godotenvallows us to read from our
.envfile we created earlier
Quick pulse check
If you’ve read our other tutorials, you’ll know at this stage we want to take our pulse. We’re a healthcare company so when we’re adding data to our blocks let’s not pick something boring like Bitcoin amounts. Put two fingers on your wrist and count how many times you feel your pulse in a minute. This is going to be your
BPM integer we’ll use throughout the tutorial.
Now let’s declare all our global variables we’ll need.
Blockis the content of each block
Blockchainis our official blockchain, that is simply a series of validated blocks. The
PrevHashin each block is compared to the
Hashof the previous block to make sure our chain is robust.
tempBlocksis simply a holding tank of blocks before one of them is picked as the winner to be added to
candidateBlocksis a channel of blocks; each node that proposes a new block sends it to this channel
announcementsis a channel where our main Go TCP server broadcasts the latest blockchain to all the nodes
mutexis a standard variable that allows us to control reads/writes and prevent data races
validatorsis a map of nodes and the amount of tokens they’ve staked
Basic blockchain functions
Before proceeding with our Proof of Stake algorithms, let’s write up our standard blockchain functions. This should be review if you’ve seen our previous tutorials. If you haven’t, that’s ok but we’ll be going through this quickly.
We start with our hashing functions.
calculateHash takes in a string and returns its SHA256 hash representation.
calculateBlockHash hashes the contents of a block by concatenating all its fields.
generateBlock is how a new block is created. The important fields we include in each new block are its hash signature (calculated by
calculateBlockHash previously) and the hash of the previous block
PrevHash(so we can keep the integrity of the chain). We also add a
Validator field so we know the winning node that forged the block.
isBlockValid performs the
PrevHash check to make sure our chain has not been corrupted.
When a validator connects to our TCP server, we need to provide it some functions that achieve a few things:
- Allow it to enter a token balance (remember for this tutorial, we won’t perform any balance checks since there is no wallet logic)
- Receive a broadcast of the latest blockchain
- Receive a broadcast of which validator in the network won the latest block
- Add itself to the overall list of validators
- Enter block data
BPM— remember, this is each validator’s pulse rate
- Propose a new block
We’ll write all this up in a
handleConn function. Here it is. Don’t worry! We’ll walk you through it all.
The first Go routine receives and prints out any announcements that come from the TCP server. These announcements will be who the winning validator is when one is chosen.
The section starting with
io.WriteString(conn, “Enter token balance:”) allows the validator to enter the number of tokens he wants to stake. Then the validator is assigned a SHA256 address and that address is added to our global
validators map we declared earlier, along with our new validator’s number of tokens staked.
We then enter
BPM which is the validator’s pulse rate and create a separate Go routine to process our block logic. The following line is important:
If the validator tries to propose a polluted block, in our case, a BPM that is not an integer, that throws an error and we immediately delete the validator from our list of validators. They are no longer eligible to forge new blocks and they lose their balance.
This potential to lose your token balance is a major reason why Proof of Stake is generally secure. If you try to alter the blockchain for your benefit and you get caught, you lose your entire staked token balance. It’s a major deterrent for bad actors.
We then create a new block with the
generateBlock function from before and we send it to the
candidateBlocks channel for further processing. Sending stuff into a channel uses this syntax:
candidateBlocks <- newBlock
for loop periodically prints the latest blockchain so each validator knows the latest state.
Picking a Winner
This is the meat of Proof of Stake logic. We need to write up how a winning validator is chosen; the higher the number of tokens they stake, the higher their probability should be to be chosen as the winner who gets to forge their block.
For simplicity in our code, we will only make validators who propose new blocks eligible to be chosen as the winner. In traditional Proof of Stake, a validator can be chosen as the winner even if they don’t propose a new block. Remember, Proof of Stake isn’t a definition, it’s a concept; there are lots of different implementations of Proof of Stake and like Proof of Work, each implementation has its own nuances.
Here is our
pickWinner function. We’ll walk you through it:
We pick a winner every 30 seconds to give time for each validator to propose a new block. Then we need to create a
lotteryPool that holds addresses of validators who could be chosen as our winner. Then we check to see there actually are some blocks proposed in our temporary holding tank of proposed blocks with
if len(temp) > 0 before proceeding with our logic.
for loop, we check to make sure we haven’t already come across the same validator in our
temp slice. If we do, skip over the block and look for the next unique validator.
In the section starting with
k, ok := setValidators[block.Validator] we make sure the validator we get from our block data in
temp is actually an eligible validator that sits in our
validators map. If they exist, then we add them to our
How do we assign proper weights to them based on the number of tokens they staked?
- We fill our
lotteryPoolwith copies of the validator’s address. They get a copy for each token they’ve staked. So a validator who put in 100 tokens will get 100 entries in the
lotteryPool. A validator who only put in 1 token will only get 1 entry.
We randomly pick the winner from our
lotteryPool and assign their address to
We then add their block to our blockchain and announce the winner to the rest of the nodes who won the lottery with this syntax, which sends the message into the
announcements <- “\nwinning validator: “ + lotteryWinner + “\n”
We clear out our
tempBlocks holding tank so it can be filled again with the next set of proposed blocks.
That is the core of a Proof of Stake consensus algorithm! That wasn’t so bad was it?
Let’s wire up our
main function now. Here it is:
We start by slurping in the contents of our
.env file which is just the port number we’ll use for our TCP server. We then create a
genesisBlock that new blocks will get added to, to form our blockchain.
We start our TCP server and expose the port in our
.env file to which new validators can connect.
We start up a Go routine to take blocks out of our
candidateBlocks channel and fill our
tempBlocks holding tank with them, for further processing by our
pickWinner function we just wrote. We then start up another Go routine for the
for loop accepts connections from new validators.
Yippee! We’re done!
Check out the full code here:
blockchain-tutorial - Write and publish your own blockchain in less than 200 lines of Go
What we just accomplished is pretty cool. We wrote up a robust Proof of Stake consensus algorithm from scratch and integrated it with actual TCP networking.
The fun stuff
Let’s try this out! Open a terminal window and start up your Go program and TCP server with
go run main.go . As expected, we get our
genesisBlock printed on our console.
Now let’s fire up a validator. Open a new terminal window and connect to our TCP server with
nc localhost 9000
We’re then prompted to add a token balance to stake. Enter the number of tokens you want that validator to stake. Then input a pulse rate for that validator.
Since we can have many validators, let’s do the same thing with another terminal window.
Watch your first terminal as you’re adding new terminals. We see validators get assigned addresses and we get a list of validators each time a new one is added!
Wait a little while and check out your new terminals. What’s happening is that our program is spending some time picking a winner. And then boom! A winner is chosen!
In our case, the first validator was chosen (we can verify the validator’s address by comparing it to the list of validators printed in our main terminal).
Wait a little while again and boom! We see our new blockchain broadcast to all our terminals, with our winning validator’s block containing his
BPM in the newest block! Cool, right?
You should be proud of getting through this tutorial. Most blockchain enthusiasts and many programmers have heard of Proof of Stake but couldn’t even begin to explain what it is. You’ve gone even further and actually built a Proof of Stake blockchain from scratch! You’re one step closer to being an expert in next generation blockchain technologies.
Because this is a tutorial, there is more we can do to make this a production ready blockchain. Things to explore next are:
- Read through our Proof of Work tutorial and tinker with it to see if you can create a hybrid blockchain
- Add time chunks where validators have a chance to propose new blocks. Our version of the code lets validators propose new blocks anytime so some blocks may periodically get cut off from consideration
- Add full peer-to-peer capability. This would basically mean each validator would run its own TCP server as well as connecting to others. We would need to add in logic so each node can find each other. Read more about this here.
Check out our other tutorials too:
- Our first, basic blockchain tutorial in less than 200 lines of Go
- How to store files through the blockchain with IPFS
- Blockchain networking (a big chunk of the tutorial you just read was based off this)
- Code your own Proof of Work
- Code a simple P2P blockchain in Go! here
- Understand concepts like Byzantine fault tolerance, Turing completeness, smart contracts and more here
- Start your own Hyperledger blockchain
- Build a DApp on Hyperledger