In May 2023, a DeFi protocol launched its token claim contract. Within 48 hours, users had spent over $2 million in gas fees, not because the contract was complex, but because the developers hadn't optimized their loops. Each claim processed the same storage array 100+ times instead of caching it once. The contract worked. It just costs users 10x more than it should have. This wasn't a hack. It wasn't a bug. It was inefficient code that silently compounded until real money started burning. Here's the reality: Compute and storage are never free. On Ethereum, every opcode has a gas price. On Stellar's Soroban, computation and storage consume transaction budgets. Solana charges compute units. Polkadot weighs transactions. Different execution models, same outcome, inefficient code becomes a production problem fast. Here's the reality: inefficient code becomes a production problem fast. If you're building on any blockchain, gas optimization isn't optional. It's the difference between a protocol users can afford to use and one they abandon for cheaper alternatives. Why Gas Costs Soar Out of Control The biggest culprit across all chains? Storage operations. Storage operations. On EVM chains (Ethereum, Polygon, BSC, Arbitrum): On EVM chains mload and mstore (memory): 3 gassload and sstore (storage): 100+ gas minimum mload and mstore (memory): 3 gas 3 gas sload and sstore (storage): 100+ gas minimum 100+ gas minimum As defined in the Ethereum Yellow Paper, storage operations cost over 100× more than memory operations. Ethereum Yellow Paper, On Non-EVM (Soroban, Solana): On Non-EVM (Soroban, Solana): Writing to a new storage entry allocates ledger space, consuming significant resourcesStorage reads/writes count against your transaction's CPU and memory budgetCross-program invocations (CPIs) add compute overhead. Writing to a new storage entry allocates ledger space, consuming significant resources Storage reads/writes count against your transaction's CPU and memory budget Cross-program invocations (CPIs) add compute overhead. The pattern is universal: Touching a persistent state is expensive. The exact mechanism differs by chain, but the principle holds. The pattern is universal: The problem isn't just that storage costs more; it's that developers unknowingly trigger these operations repeatedly. A loop that runs 100 times and reads from storage on each iteration? That's 100 expensive reads instead of one cached value, whether you're on EVM or Non-EVM. Small inefficiencies multiply. And by the time you notice in production, your users are already paying for it. The 6 Gas Killers (And How to Fix Them) Most gas bloat comes from patterns developers don't recognize as problems until deployment. These patterns appear across all blockchain architectures; the syntax changes, but the inefficiencies remain the same. all blockchain architectures Let's break down the worst offenders with examples across different chains: 1. Unnecessary Storage Writes. The mistake: Writing default or zero values when they're not needed, or writing data that hasn't actually changed. The mistake: Why it's expensive: EVM chains: Each storage write costs ~20,000 gas. Even after EIP-3529 reduced refunds, writing to an already-zero slot still costs ~2,900 gas on warm slots.Non-EVM chains (e.g., Soroban): Every storage write consumes part of your transaction budget. Writing unchanged values wastes CPU instructions and ledger I/O. EVM chains: Each storage write costs ~20,000 gas. Even after EIP-3529 reduced refunds, writing to an already-zero slot still costs ~2,900 gas on warm slots. EVM chains: Non-EVM chains (e.g., Soroban): Every storage write consumes part of your transaction budget. Writing unchanged values wastes CPU instructions and ledger I/O. Non-EVM chains (e.g., Soroban): How to fix it: On EVM (Solidity): On EVM (Solidity): // ❌ Bad: Always writes
balances[user] = 0;

// ✅ Good: Only writes if needed
if (balances[user] != 0) {
    delete balances[user]; // Triggers gas refund
} // ❌ Bad: Always writes
balances[user] = 0;

// ✅ Good: Only writes if needed
if (balances[user] != 0) {
    delete balances[user]; // Triggers gas refund
} On Non-EVM (Rust/Soroban): On Non-EVM (Rust/Soroban): // ❌ Bad: Always writes
env.storage().instance().set(&USER_BALANCE, &0);

// ✅ Good: Check before writing
if env.storage().instance().has(&USER_BALANCE) {
    env.storage().instance().remove(&USER_BALANCE);
} // ❌ Bad: Always writes
env.storage().instance().set(&USER_BALANCE, &0);

// ✅ Good: Check before writing
if env.storage().instance().has(&USER_BALANCE) {
    env.storage().instance().remove(&USER_BALANCE);
} Additional optimizations: EVM: Pack booleans and small integers into the same 256-bit slot, save entire storage slots (~20,000 gas each).Non-EVM (Soroban): Use temporary storage for data that doesn't need to persist across ledgers. Both: Use events/logs instead of storage for infrequently accessed data. EVM: Pack booleans and small integers into the same 256-bit slot, save entire storage slots (~20,000 gas each). EVM: Non-EVM (Soroban): Use temporary storage for data that doesn't need to persist across ledgers. Non-EVM (Soroban): Both: Use events/logs instead of storage for infrequently accessed data. Both: 2. Loops That Access Storage Repeatedly The mistake: Reading from or writing to storage inside loops. Each storage operation costs resources, so if your loop touches storage on every iteration, costs multiply quickly. The mistake: Why it's expensive: A loop that processes 100 items and reads from storage each time equals 100 expensive operations instead of 1 cached read. How to fix it: On EVM (Solidity): On EVM (Solidity): // ❌ Bad: Reads storage 100 times
for (uint i = 0; i < users.length; i++) {
    totalBalance += balances[users[i]]; // sload every iteration
}

// ✅ Good: Cache values
uint length = users.length;
uint tempBalance;
for (uint i = 0; i < length; i++) {
    tempBalance += balances[users[i]];
}
totalBalance = tempBalance; // Single sstore // ❌ Bad: Reads storage 100 times
for (uint i = 0; i < users.length; i++) {
    totalBalance += balances[users[i]]; // sload every iteration
}

// ✅ Good: Cache values
uint length = users.length;
uint tempBalance;
for (uint i = 0; i < length; i++) {
    tempBalance += balances[users[i]];
}
totalBalance = tempBalance; // Single sstore On Non-EVM (Rust/Soroban): On Non-EVM (Rust/Soroban): // ❌ Bad: Reads storage in loop
let mut total = 0;
for user in users.iter() {
    total += env.storage().instance().get::<_, i128>(&user).unwrap_or(0);
}


// ✅ Good: Batch reads or cache
let total: i128 = users.iter()
    .map(|u| env.storage().instance().get::<_, i128>(u).unwrap_or(0))
    .sum(); // ❌ Bad: Reads storage in loop
let mut total = 0;
for user in users.iter() {
    total += env.storage().instance().get::<_, i128>(&user).unwrap_or(0);
}


// ✅ Good: Batch reads or cache
let total: i128 = users.iter()
    .map(|u| env.storage().instance().get::<_, i128>(u).unwrap_or(0))
    .sum(); 3. Not Validating Inputs Early The mistake: Performing expensive operations before checking if inputs are even valid. The mistake: Why it's expensive: If a transaction is going to fail anyway, you want it to fail as cheaply as possible. Order matters. How to fix it: On EVM (Solidity) On EVM (Solidity) // ❌ Bad: Expensive check first
require(balances[msg.sender] >= amount); // Storage read
require(amount > 0); // Memory check

// ✅ Good: Cheap checks first
require(amount > 0); // Fails fast if zero
require(balances[msg.sender] >= amount); // Only runs if amount valid // ❌ Bad: Expensive check first
require(balances[msg.sender] >= amount); // Storage read
require(amount > 0); // Memory check

// ✅ Good: Cheap checks first
require(amount > 0); // Fails fast if zero
require(balances[msg.sender] >= amount); // Only runs if amount valid On Non-EVM (Rust/Soroban): On Non-EVM (Rust/Soroban): // ❌ Bad: Storage check first
let balance = env.storage().instance().get::<_, i128>(&user).unwrap_or(0);
if amount <= 0 {
    panic!("Invalid amount");
}

// ✅ Good: Validate inputs first
if amount <= 0 {
    panic!("Invalid amount"); // Fails immediately
}
let balance = env.storage().instance().get::<_, i128>(&user).unwrap_or(0);
if balance < amount {
    panic!("Insufficient balance");
} // ❌ Bad: Storage check first
let balance = env.storage().instance().get::<_, i128>(&user).unwrap_or(0);
if amount <= 0 {
    panic!("Invalid amount");
}

// ✅ Good: Validate inputs first
if amount <= 0 {
    panic!("Invalid amount"); // Fails immediately
}
let balance = env.storage().instance().get::<_, i128>(&user).unwrap_or(0);
if balance < amount {
    panic!("Insufficient balance");
} Universal principle: Validate cheaply (bounds checks, null checks) before expensive operations (storage reads, cryptographic operations). Universal principle: 4. Making Multiple Transactions Instead of Batching. The mistake: Splitting work across many transactions instead of grouping operations. The mistake: Why it's expensive: EVM chains: Each transaction has a base fee (~21,000 gas). 100 transactions = 100 base fees. EVM chains: Non-EVM (eg, Soroban): Each transaction consumes base CPU/memory budget overhead. Non-EVM (eg, Soroban): How to fix it: On EVM (Solidity): On EVM (Solidity): // ❌ Bad: 100 transactions = 100 base fees
// User calls transfer() 100 times

// ✅ Good: 1 transaction, 1 base fee
function batchTransfer(address[] calldata recipients, uint[] calldata amounts) external {
    require(recipients.length == amounts.length, "Length mismatch");
    for (uint i = 0; i < recipients.length; i++) {
        _transfer(msg.sender, recipients[i], amounts[i]);
    }
} // ❌ Bad: 100 transactions = 100 base fees
// User calls transfer() 100 times

// ✅ Good: 1 transaction, 1 base fee
function batchTransfer(address[] calldata recipients, uint[] calldata amounts) external {
    require(recipients.length == amounts.length, "Length mismatch");
    for (uint i = 0; i < recipients.length; i++) {
        _transfer(msg.sender, recipients[i], amounts[i]);
    }
} On Non-EVM (Soroban/Rust): On Non-EVM (Soroban/Rust): // ✅ Good: Batch operations in single contract call
pub fn batch_transfer(
    env: Env,
    from: Address,
    transfers: Vec<(Address, i128)>
) -> Result<(), Error> {
    from.require_auth();
    
    for (to, amount) in transfers.iter() {
        transfer_internal(&env, &from, to, *amount)?;
    }
    Ok(())
} // ✅ Good: Batch operations in single contract call
pub fn batch_transfer(
    env: Env,
    from: Address,
    transfers: Vec<(Address, i128)>
) -> Result<(), Error> {
    from.require_auth();
    
    for (to, amount) in transfers.iter() {
        transfer_internal(&env, &from, to, *amount)?;
    }
    Ok(())
} 5. Inefficient Data Handling The mistake: Copying data unnecessarily or using the wrong data location for function parameters. The mistake: Why it's expensive: Memory operations and data copies cost resources across all chains. How to fix it: On EVM (Solidity): On EVM (Solidity): // ❌ Bad: Copies array to memory
function sum(uint[] memory numbers) public returns (uint) {
    uint total;
    for (uint i = 0; i < numbers.length; i++) {
        total += numbers[i];
    }
    return total;
}

// ✅ Good: Reads directly from calldata
function sum(uint[] calldata numbers) public pure returns (uint) {
    uint total;
    for (uint i = 0; i < numbers.length; i++) {
        total += numbers[i];
    }
    return total;
} // ❌ Bad: Copies array to memory
function sum(uint[] memory numbers) public returns (uint) {
    uint total;
    for (uint i = 0; i < numbers.length; i++) {
        total += numbers[i];
    }
    return total;
}

// ✅ Good: Reads directly from calldata
function sum(uint[] calldata numbers) public pure returns (uint) {
    uint total;
    for (uint i = 0; i < numbers.length; i++) {
        total += numbers[i];
    }
    return total;
} On Non-Evm (Soroban/Rust): On Non-Evm (Soroban/Rust): // ❌ Bad: Cloning data unnecessarily
pub fn process_data(env: Env, data: Vec<i128>) -> i128 {
    let copied = data.clone(); // Unnecessary clone
    copied.iter().sum()
}

// ✅ Good: Use references
pub fn process_data(env: Env, data: Vec<i128>) -> i128 {
    data.iter().sum() // No clone needed
} // ❌ Bad: Cloning data unnecessarily
pub fn process_data(env: Env, data: Vec<i128>) -> i128 {
    let copied = data.clone(); // Unnecessary clone
    copied.iter().sum()
}

// ✅ Good: Use references
pub fn process_data(env: Env, data: Vec<i128>) -> i128 {
    data.iter().sum() // No clone needed
} Additional tips: EVM (Solidity): Mark read-only functions as view or pure (can be called off-chain for free).Non-EVM (Rust): Use `&` references instead of cloning Vec or Map structures. EVM (Solidity): Mark read-only functions as view or pure (can be called off-chain for free). EVM (Solidity): Non-EVM (Rust): Use `&` references instead of cloning Vec or Map structures. Non-EVM (Rust): & Your Gas Optimization Checklist Before you deploy on any chain, run through this checklist: any chain Storage usage: Storage usage: No unnecessary writes to storageVariables/data packed efficientlyOld entries deleted/removed for refunds (EVM) or budget recoveryEvents/logs are used instead of storage for infrequently accessed data No unnecessary writes to storage Variables/data packed efficiently Old entries deleted/removed for refunds (EVM) or budget recovery Events/logs are used instead of storage for infrequently accessed data Loops: Loops: No storage reads/writes inside loopsArray lengths and values are cached beforehandHeavy computation moved off-chain or batched No storage reads/writes inside loops Array lengths and values are cached beforehand Heavy computation moved off-chain or batched Data locations: Data locations: EVM (Solidity): Large inputs use calldata, not memory.Non-Evm (Rust): Use references (&) instead of cloning. EVM (Solidity): Large inputs use calldata, not memory. EVM (Solidity): Non-Evm (Rust): Use references (&) instead of cloning. Non-Evm (Rust): Control flow: Control flow: Cheap validation checks before expensive operationsConditions are structured to short-circuit away from costly paths. Cheap validation checks before expensive operations Conditions are structured to short-circuit away from costly paths. Redundant operations: Redundant operations: Current values checked before writingConditional checks skip no-op writesState is cleared explicitly when no longer needed Current values checked before writing Conditional checks skip no-op writes State is cleared explicitly when no longer needed Profiling: Profiling: Gas/resource reporter integrated into test suite.Functions profiled during development.Static analysis in the CI/CD pipeline. Gas/resource reporter integrated into test suite. Functions profiled during development. Static analysis in the CI/CD pipeline. Chain-specific optimizations: Chain-specific optimizations: EVM: Storage packing, immutable/constant usage, library calls.Non-EVM (Rust): Storage tier usage, WASM size, Rust efficiency patterns. EVM: Storage packing, immutable/constant usage, library calls. EVM: Non-EVM (Rust): Storage tier usage, WASM size, Rust efficiency patterns. Non-EVM (Rust): Alright, real talk Gas optimization isn't glamorous. It doesn't lend itself to flashy demos or impressive pitch decks. But it's the difference between a protocol users can actually afford to use and one they abandon after the first transaction. I've seen great projects die because their gas costs were five times higher than those of their competitors. Not because the code was bad, but because it wasn't efficient. efficient Every blockchain has its quirks, but the fundamentals are universal: plan for efficiency or pay the price. Every blockchain has its quirks, but the fundamentals are universal: plan for efficiency or pay the price. Whether you're deploying to Ethereum mainnet, launching on Stellar, or building on any other Non-EVM chains, the same patterns apply: minimize storage, batch operations, validate early, and compute off-chain. minimize storage, batch operations, validate early, and compute off-chain.

This story contains AI-generated text. The author has used AI either for research, to generate outlines, or write the text itself. 

The is an opinion piece based on the author’s POV and does not necessarily reflect the views of HackerNoon.

Gas Inefficiencies Developers Don't Notice Until It's Too Late

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Researchers Rethink Masking in Diffusion-Based Text Generation

Learn Technical Storytelling This Halloween at the Spells & Syntax Workshop

Why ‘Good’ Performance Metrics Might Be Killing Your Business

An App for Tracking Gas Fee and Save Your Money

Explaining ETH Gas and How It Impacts Ethereum Network

Ethereum Dencun: Transactions Fees Drop by More Than 90%

Researchers Rethink Masking in Diffusion-Based Text Generation

Learn Technical Storytelling This Halloween at the Spells & Syntax Workshop

Why ‘Good’ Performance Metrics Might Be Killing Your Business

An App for Tracking Gas Fee and Save Your Money

Explaining ETH Gas and How It Impacts Ethereum Network

Ethereum Dencun: Transactions Fees Drop by More Than 90%

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps