How does UCall compare to FastAPI for microservices?

UCall is significantly faster, with benchmarks showing up to 100x lower latency and higher throughput, especially using io_uring. However, FastAPI offers a more mature ecosystem, HTTPS out-of-the-box, and broader community support, making it better for general-purpose web apps.

How to set up a UCall server in Python?

Install via pip with 'pip install ucall', then import Server from ucall.uring (for Linux 5.19+) or ucall.posix. Decorate functions with @server and run server.run(). The README provides a simple sum function example similar to FastAPI but with performance optimizations.

Does UCall support HTTPS for secure connections?

Not currently. HTTPS support is listed on the roadmap as a future feature, so for now, UCall is limited to HTTP over TCP, which may not be suitable for production environments requiring encryption.

What Linux kernel version is needed for UCall's best performance?

You need Linux kernel 5.19 or newer to fully utilize io_uring features like direct accept and sparse file registration. Older kernels can use the POSIX backend, but with reduced performance, as shown in the benchmark tables.

Can UCall handle image uploads in AI applications?

Yes, UCall natively serializes PIL images and NumPy arrays, making it efficient for AI models. The README includes an example with UForm for vectorizing images and text, leveraging this rich type support without extra serialization overhead.

Is UCall stable for production deployment?

While UCall offers extreme performance, it's a newer project with incomplete features like HTTPS and a smaller community. It's best suited for performance-critical, Linux-based environments where these trade-offs are acceptable, but caution is advised for broad production use.

Open-Awesome

UCall

Apache-2.0Cv0.5.8

A high-performance JSON-RPC and REST library for Python and C, leveraging io_uring and SIMD for ultra-low latency and high throughput.

Visit Website GitHub

1.3k stars54 forks0 contributors

What is UCall?

UCall is a remote procedure call library that implements JSON-RPC and REST over TCP/HTTP with a focus on extreme performance. It solves the problem of high latency and low throughput in traditional web serving libraries by leveraging Linux io_uring and SIMD-accelerated parsers, making it ideal for high-demand microservices and AI applications.

Target Audience

Developers building high-performance microservices, real-time AI inference pipelines, or any system where low-latency RPC is critical, especially those deploying on Linux with kernel 5.19+.

Value Proposition

Developers choose UCall for its unmatched performance—up to 100x faster than FastAPI—and its ability to handle custom data types like NumPy arrays natively, all while maintaining a simple API comparable to mainstream frameworks.

Overview

Web Serving and Remote Procedure Calls at 50x lower latency and 70x higher bandwidth than FastAPI, implementing JSON-RPC & REST over io_uring ☎️

Use Cases

Best For

Building ultra-low-latency microservices requiring high request throughput
Deploying multi-modal AI models that need efficient image and text embedding serialization
Replacing FastAPI or gRPC in performance-critical backend services
Running efficient JSON-RPC servers on free-tier cloud instances
Developing real-time data processing pipelines with minimal network overhead
Integrating high-performance RPC into existing C or Python applications

Not Ideal For

Teams deploying on non-Linux systems or older Linux kernels without io_uring support
Projects requiring built-in HTTPS encryption for secure production deployments
Developers needing a mature ecosystem with extensive third-party middleware and plugins

Pros & Cons

Pros

Ultra-Low Latency

Leverages io_uring and SIMD parsers to achieve round-trip times as low as 22 microseconds, significantly outperforming FastAPI and gRPC in benchmarks.

High Throughput

Handles up to 231,000 requests per second on supported hardware, making it ideal for high-demand microservices and AI inference pipelines.

Rich Type Serialization

Natively supports custom types like NumPy arrays and PIL images, facilitating efficient data handling for multi-modal AI applications, as shown in the UForm integration example.

Cost-Effective Deployment

Delivers high performance on free-tier cloud instances like AWS t4g.small, with throughput up to 88,455 RPS, reducing operational costs.

Cons

Linux Kernel Dependency

Optimal performance requires Linux kernel 5.19+ for io_uring features, excluding Windows, macOS, and older Linux systems, limiting cross-platform use.

Missing Production Features

Critical features like HTTPS support and WebSockets are still on the roadmap, not yet available, which may hinder adoption in secure or real-time web environments.

Verbose C API

The C interface requires manual argument extraction with functions like ucall_param_named_i64, making it less ergonomic and more error-prone compared to higher-level libraries.

Open Source Alternative To

UCall is an open-source alternative to the following products:

gRPC

FastAPI

A modern, fast web framework for building APIs with Python 3.7+ based on standard Python type hints, with automatic OpenAPI documentation.

Frequently Asked Questions

Related Projects

uWebSockets

Simple, secure & standards compliant web server for the most demanding of applications

Stars18,959

Forks1,860

Last commit3 days ago

KCP

:zap: KCP - A Fast and Reliable ARQ Protocol

Stars16,875

Forks2,626

Last commit2 months ago

Cpp-HttpLib

A C++ header-only HTTP/HTTPS server and client library

Stars16,757

Forks2,739

Last commit2 days ago

Muduo

Event-driven network library for multi-threaded Linux server in C++11

Stars16,227

Forks5,324

Last commit8 months ago

Community-curated · Updated weekly · 100% open source

Found a gem we're missing?

Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.

Submit a project Star on GitHub