arXiv:2505.04129
arXiv Year 2025 Preprint
Blockchain · Performance

Maxing Out the SVM: Performance Impact of Memory and Program Cache Sizes in the Agave Validator

Turan Vural Yuki Yuminaga Alex Petrosyan Benjamin Livshits
2025
Publication year
arXiv
Venue
Preprint
Type

Problem

In this paper we analyze some of the bottlenecks in the execution pipeline of Solana's Agave validator client, focusing on RAM and program cache usage under mainnet conditions.

Approach

Through a series of controlled experiments, we measure the validator's throughput and resource efficiency as RAM availability ranges between 128 GB to 1,536 GB (1.5 TB).

Results

We discover that the validator performance degrades significantly below 256 GB, with transaction processing falling behind real-time block production. Additionally, we study the program cache behavior, identifying inefficiencies in program eviction and load latency. Our results provide practical guidance for hardware provisioning and suggest improvements to the Solana execution and caching strategy, reducing latency due to the program cache by 90%.

Cite this paper — BibTeX
@TechReport{vural25maxing,
  title = "{Maxing Out the SVM: Performance Impact of Memory and Program Cache Sizes in the Agave Validator}",
  author = "Turan Vural and Yuki Yuminaga and Alex Petrosyan and Benjamin Livshits",
  year = "2025",
  month = may,
  institution = "arXiv",
  number = "arXiv:2505.04129",
}
Copied