Bilingual

Show HN: Simulate API traffic directly from your IDE with .http files
20 days ago
- Modern architecture designed for massive scalability, surpassing legacy tools that rely on heavy OS threads.
- Utilizes a pure-Go lock-free Actor Model to isolate connections into ultra-lightweight Goroutines.
- Achieves over 30,000 RPS with zero-garbage overhead (0 allocs/op) in a single instance.
Distributed System Is Slower Than a Laptop
15 days ago
- A 2015 paper showed single-threaded laptop outperformed distributed systems on large-scale graph processing tasks.
- Many distributed systems fail to beat a single competent thread even with hundreds of cores, leading to high COST (Configuration that Outperforms a Single Thread).
- Companies pay millions for distributed systems when simpler, cheaper single-machine solutions suffice and perform better.
- Hardware improvements have made single servers extremely powerful, but distributed-by-default reflexes persist without justification.
- Distributed systems often introduce unnecessary complexity, higher costs (compute, engineering, incidents), and slower performance.
- Frameworks, career incentives, vendor pricing, and misconceptions about scaling drive unnecessary distribution.
- Legitimate reasons for distribution include data size, availability needs, geographic latency, and organizational partitioning.
- Building a single-machine baseline before adopting distributed architectures can save costs and improve design decisions.
Scaling to 1M concurrent sandboxes in seconds
5 days ago
- Modal has rebuilt its sandbox platform to handle massive scale, enabling millions of concurrent sandboxes and creation rates of tens of thousands per second.
- Traditional container platforms like Kubernetes face scaling bottlenecks from O(containers) or O(nodes) operations, such as centralized scheduling and etcd writes.
- Modal's new architecture eliminates central coordination, using a fleet of scheduling servers and workers as independent sources of truth, with no durable stores in the critical path.
- Benchmarks show the system can create 1 million sandboxes in under a minute, with median start times under half a second and no significant degradation at scale.
- The development involved a focused prototype effort and required reimplementing features and optimizing container startup to handle high concurrency.
Zenoh Reliability, Scalability and Congestion Control
9 hours ago
- Reliable many-to-many communication over wide area networks is challenging, requiring trade-offs between reliability, scalability, and resource consumption.
- zenoh's protocol has two layers: a session protocol for 1-to-1 communication with best-effort and reliable channels, and a routing protocol to propagate data from producers to consumers.
- Hop-to-hop reliability is scalable and reliable under stable infrastructure but may lose data during topology changes.
- Stronger reliability options include end-to-end (no loss during topology changes but less scalable) and first-router-to-last-router (reduces pressure on endpoints).
- Flow control separates control over resending (receiver), memory usage (sender/infrastructure), and message dropping (sender congestion control).
- Configuration examples show reliability set on the subscriber side and congestion control per sample using the Python API.
- zenoh allows balancing application needs with system constraints, protecting nodes from asymmetries and maintaining strong guarantees in dynamic networks.

Hasty Briefsbeta

#scalability

Show HN: Simulate API traffic directly from your IDE with .http files

Distributed System Is Slower Than a Laptop

Scaling to 1M concurrent sandboxes in seconds

Zenoh Reliability, Scalability and Congestion Control