The Aptos Challenge: Web3 Throughput and Latency Innovation in Support of 5 Billion Internet Users

Author:Avery Ching, Founder of Aptos

As Web3 technologies evolve, the infrastructure that supports billions of users faces unprecedented challenges.Aptos, a program dedicated to building efficient, reliable and scalable networks, has achieved significant results in improving performance. This paper explores how Aptos is addressing the upcoming challenges, specifically how dynamic parallelism can improve the user experience.

current state

Aptos recently deployed a theoretically optimal consensus latency solution (i.e., AIP-ordered voting) on its main network, achieving consistently less than one-second user latency under low to moderate workloads. The impact of this advancement can be tracked in real-time at the following link:Aptos Labs GrafanaThe

Additionally, Aptos set a record for processing 154.4 million transactions in a single day and demonstrated a daily processing capacity of more than 2 billion transactions in a mainnet-like environment, achievements that mark new milestones for the industry.

Challenges faced

Despite the remarkable achievements that have been made, Aptos still faces two major challenges:

1. Improved user latency under high workloads: User latency in high volume environments (e.g., 20,000 TPS) needs to be further reduced to support applications such as@Visarespond in singing@Nasdaqapplication scenarios for large organizations such as
2. Continue to advance concurrency and gas prioritization: There is a need to ensure that non-conflicting transactions can run in parallel, while rationally managing gas costs to optimize resource allocation.

Solutions for Dynamic Parallelism

To overcome these challenges, Aptos has adopted a dynamic parallelism approach based on two key technologies: BlockSTM and a pipelined and modular architecture.

• BlockSTM: By extracting runtime concurrency from all transactions, BlockSTM can significantly improve processing efficiency.
• Streamlining and Modular Architecture: Allows to hide delays between stages and to easily integrate new technologies such as the Shoal++ consensus protocol.

By combining these technologies, Aptos is able to achieve the following goals:

• Runtime Tip: Optimize concurrent scheduling using information such as addresses, developer-specified dependencies, etc.
• run-time concurrency: Enables operations that appear to require serial processing (e.g., order book processing) to be executed in parallel.
• Dynamic resource optimization: Dynamically adjust core utilization based on transaction type and resource consumption to ensure efficient processing.
• feedback loop: Dynamically optimize resource usage by collecting historical data to achieve self-optimization of the system.

reach a verdict

While these technical challenges are complex and daunting, the Aptos team is working to solve them and will continue to share progress. By adopting a strategy of dynamic parallelism, Aptos aims to build a robust, decentralized infrastructure for all. We look forward to working with community members to meet these challenges and shape the future of Web3.