On a 1 gbps network connections, the maximum number of transactions is 710 k transactions per second. Due to Ethernet framing, a loss of 1-4% is expected. 

Memory Access Pattern Efficiency

The leader is expected to take in incoming user packets and efficiently order them into a proof of history sequence, which is then published to downstream verifiers. Efficiency is determined based on the memory access patterns of the transactions, so transactions are ordered to minimize faults and maximize prefetching. 

University of Connecticut

- Components
- Network Limits
- Computational Limits
- Memory Limits
- High Performance Smart Contracts 

Solana - System Architecture

Solana - A new architecture for a high performance blockchain

- Leader and Proof of History
- State
- Verifier and State Replication
- Validators

Components

Network Limits

Each transaction needs a digest verification. Because it does not use any memory outside the transaction message itself, the operation can be parallelized independently. Output is therefore limited by the number of cores in the system. 

Computational Limits

A naïve implementation of the state would theoretically fit 10 billion accounts into 640 GB. Based on 2 reads and 2 writes per transaction, memory throughput can handle 2.75 million transactions per second.

Memory Limits

Smart contracts are a generalized form of transactions. They are programs running on each node to modify the state. One of its advantages is a zero cost foreign function interface intrinsic, which are functions implemented on the platform directly and therefore callable by programs. Calling the intrinsic suspends that program and schedules the intrinsic on a higher performance server. 

Learn Before

Related

Learn After