⬅️

matt_d 11 hoursReload

A Study of Undefined Behavior Across Foreign Function Boundaries in Rust Libs

matt_d 19 hoursReload

Banked Memories for Soft SIMT Processors

matt_d 6 daysReload

Analyzing Modern NVIDIA GPU cores

matt_d 8 daysReload

Abstract: https://arxiv.org/abs/2405.11182

PDF: https://arxiv.org/pdf/2405.11182

Section 8, Conclusion:

    In this paper, we quantify the overhead of running a state machine replication system for cloud systems written in a language with GC. To this end, we (1) design from scratch a canonical cloud system—a distributed, consensus-based, linearizable key-values store, (2) implement it in C++, Java, Rust, and Go, (3) evaluate implementations under an update-heavy and read-heavy workloads on AWS under different resource constraints while trying to hit the maximum throughput with a fixed low tail latency. Our results show that even with ample memory, GC has a non-trivial cost, and with limited memory, languages with memory management can achieve an order of magnitude higher throughput than the languages with GC on the same hardware. Our key observation is that if a cloud system is expected to grow to a large volume of users, building the system in a language with manual memory management and thereby paying a higher development cost than using a language with GC may result in a significant cloud cost savings in the long run.

matt_d 8 daysReload

Ratte: Fuzzing Multi-Level Compilers Using Composable Semantics [pdf]

matt_d 8 daysReload

Abstract: https://arxiv.org/abs/2405.11182

PDF: https://arxiv.org/pdf/2405.11182

Section 8, Conclusion:

    In this paper, we quantify the overhead of running a state machine replication system for cloud systems written in a language with GC. To this end, we (1) design from scratch a canonical cloud system—a distributed, consensus-based, linearizable key-values store, (2) implement it in C++, Java, Rust, and Go, (3) evaluate implementations under an update-heavy and read-heavy workloads on AWS under different resource constraints while trying to hit the maximum throughput with a fixed low tail latency. Our results show that even with ample memory, GC has a non-trivial cost, and with limited memory, languages with memory management can achieve an order of magnitude higher throughput than the languages with GC on the same hardware. Our key observation is that if a cloud system is expected to grow to a large volume of users, building the system in a language with manual memory management and thereby paying a higher development cost than using a language with GC may result in a significant cloud cost savings in the long run.