Hyperloop: An Integrated Release Scheduling System for Meituan iOS Clients

What is Hyperloop? Hyperloop is a release‑scheduling system used by Meituan‑Dianping for iOS client component publishing, continuous integration, app packaging, and resource allocation. Its name is inspired by Elon Musk’s Hyperloop high‑speed rail , symbolizing modernity, simplicity, and efficiency.

Hyperloop provides a one‑stop platform that manages more than 300 iOS components and four Meituan‑Dianping apps. After integration, developers configure release steps and checks in Hyperloop; a single login allows them to trigger the configured workflow, with real‑time status reporting, error logs, and notifications.

Why Hyperloop? The previous Jenkins‑based workflow suffered from several issues as the business grew:

1. Increased usage and maintenance cost

More developers participated in the release process, but Jenkins required professional configuration and operational knowledge, leading to heavy maintenance overhead for ordinary developers.

2. Explosion of Jenkins jobs

The number of Jenkins jobs grew dramatically. A snapshot shows 74 top‑level directories and 569 total jobs.

Top‑level directories

Total jobs

74

569

Each business unit duplicated example jobs and customized them, causing a combinatorial explosion of tasks and a proportional increase in maintenance effort.

3. Difficulty aggregating build data

Jenkins provides weak data analysis and presentation capabilities, making it hard to collect, store, and visualize build metrics such as warning counts. The lack of centralized data also hindered timely feedback and decision‑making.

These problems motivated the design of a centralized scheduling system—Hyperloop.

How to build Hyperloop?

1. System Architecture

Hyperloop is divided into three layers: a web front‑end for user interaction and data visualization, a back‑end that handles data collection, business logic, and task scheduling, and a toolchain that executes the actual release and integration steps.

The back‑end receives actions from the front‑end, invokes the toolchain, and ensures that all task data flow through it, guaranteeing data integrity.

2. Business Model

Plugin‑based checks

Release and integration steps are modularized as plugins. Each component can select required plugins, and optional plugins can be enabled based on business needs. Some checks (e.g., warning thresholds) allow target values and timelines, enabling dynamic calculation of required metrics.

Capabilities

The toolchain provides Fastlane‑style actions that can be parameterized. After development, capabilities are published in the Hyperloop back‑end and become selectable in component settings.

Strategies

When a capability is selected, a corresponding strategy is generated, storing the component‑specific parameters.

Steps

Each release or integration creates a version; the back‑end generates concrete execution steps based on selected capabilities and strategies. The toolchain receives these steps and executes them with the provided parameters.

3. Feature Implementation

Component Release

More than 300 components and over 700 release requests are managed by Hyperloop. The first phase delivered basic component publishing; subsequent phases added optional optimization capabilities such as warning‑count analysis and limits.

Dynamic target values can be set for metrics, and the system automatically calculates the required thresholds for each release.

Additional features such as Changelog generation and detailed error logs improve visibility of each release.

Packaging & Integration

After component release, the system supports integration into the Meituan iOS client and other apps. The integration pipeline includes size checks, automated tests, and full source compilation to ensure API compatibility.

Build artifacts are uploaded to Meituan Cloud with QR‑code download links for developers and QA.

For smaller business units that only need packaging, Hyperloop provides templated build configurations (e.g., feature tests, daily builds, App Store submissions) that can be triggered with a single click.

Jenkins De‑coupling

Initially Hyperloop still used Jenkins for task dispatch, but the back‑end and toolchain are both Ruby‑based, allowing direct message‑queue communication and database access. This eliminated the need for a separate Jenkins instance.

Notification Capability

Hyperloop sends real‑time notifications via the internal IM tool (Elephant) and email. Users receive immediate alerts for failures, as well as success messages, certificate expirations, provisioning profile updates, and other important news.

Can Hyperloop solve the problems?

1. Replaces Jenkins with a modern toolchain and back‑end, reducing learning and operational costs.

2. Provides a centralized system, eliminating fragmented configurations and lowering maintenance overhead.

3. Emphasizes data collection and analysis, enabling metric‑driven optimization.

Future Outlook

1. Monitoring & Statistics

Expand monitoring to include static analysis, duplicate code detection, and other quality metrics.

2. Upstream/Downstream Integration

Connect Hyperloop with Git platforms, Meituan Cloud, iTC, IM, and other systems to automate reporting, task creation, and wiki updates.

3. Multi‑Platform Support

Extend beyond iOS to support Android and other platforms, making the system a universal release engine.

4. Process Optimization

Continuously refine workflows to improve build efficiency, leveraging modern production‑relationship theory to align tooling with developer needs.

Author Bio Ensheng (@zesming) is a senior engineer at Meituan‑Dianping. He joined the original Meituan in 2014, worked on core iOS features, and now leads iOS release engineering, designing Hyperloop and developing its back‑end.