Popular：
Virtualization DNS security formal verification reachability analysis compiler errors macro conflict web extension development framework Bitmap Graphics API inconsistencies All Tags

Codon Compiler: A Glimpse of Faster Python?

2025-03-16
Codon Compiler: A Glimpse of Faster Python?

Codon is a compiler aiming to dramatically improve Python's execution speed. While the author previously encountered compilation issues, recent updates have resolved them. Although a test script didn't show speed improvements, Codon demonstrated significant performance gains in NPBench NumPy benchmarks, reaching up to 900x speedup, largely due to the Codon team's direct port of NumPy. While the author didn't replicate the benchmark results, a trial of a supposedly 300x faster Python script hints at Codon's potential in specific use cases.

Read more
(www.usenix.org)
Development

NSDI '24: Autothrottle: A Practical Bi-Level Approach to Resource Management for SLO-Targeted Microservices

2025-02-03

USENIX is committed to Open Access, making research from its events freely available. Papers, proceedings, and any subsequent video/audio/slides are open to all after the event. This includes the NSDI '24 paper, "Autothrottle: A Practical Bi-Level Approach to Resource Management for SLO-Targeted Microservices," by Wang et al., presenting a practical approach to managing resources for SLO-targeted microservices. The paper, video, and slides are now publicly accessible.

Read more
(www.usenix.org)
Development

Google SRE's Evolution: From Error Budgets to Systems Theory

2025-01-03
Google SRE's Evolution: From Error Budgets to Systems Theory

Google's Site Reliability Engineering (SRE) team has undergone a significant evolution over the past 25 years. Initially relying on methods like Service Level Objectives (SLOs), error budgets, and isolation strategies, Google's SRE team has shifted towards systems theory and control theory, adopting the STAMP framework to address increasingly complex systems and emerging challenges. STAMP shifts the focus from preventing individual component failures to understanding and managing complex system interactions. This article uses a real-world case study to illustrate how STAMP helps Google prevent system-level failures and explores its future applications across the tech industry.

Read more
(www.usenix.org)
Development Systems Theory

Google's New Load Balancer, PReQuaL: Beyond CPU Load Balancing

2024-12-16

Google Research presented PReQuaL (Probing to Reduce Queuing and Latency), a novel load balancer, at NSDI 2024. Unlike traditional CPU load balancing, PReQuaL actively probes server latency and active requests to select servers, dramatically reducing tail latency, error rates, and resource consumption in systems like YouTube. Deployed in YouTube for over a year, PReQuaL has significantly improved system utilization. This innovative approach challenges conventional wisdom and offers a new paradigm for high-performance distributed systems.

Read more
(www.usenix.org)
Development load balancing