Seastar: new C++ framework for web-scale workloads
Blog post from ScyllaDB
Seastar is a newly released open-source C++ framework designed by the ScyllaDB team to provide extreme high-performance capabilities for web-scale workloads on OSv and Linux, offering a 5x throughput improvement and handling millions of transactions per second on a single server. It targets modern physical and virtual hardware by leveraging the shard-per-core approach, which enables linear scalability by avoiding synchronous data exchange and locking across cores. Instead, each core operates independently with its own resources and communicates asynchronously with others, aided by a user-space network stack based on the Data Plane Development Kit (DPDK) to eliminate system calls and data copying. Seastar facilitates the development of high-throughput, low-latency network applications such as NoSQL servers, data caches, and high-performance HTTP serving by enabling non-blocking, asynchronous server code that is easier to debug and optimize. It is available under the Apache license version 2.0, with resources and updates accessible via the Seastar project site and ScyllaDB's social media channels.
No tracked trend matches for this post yet.