Glossary

Glossary#

This section introduces commonly used terminology in openYuanrong.

openYuanrong Function: The basic unit of distributed scheduling and operation in openYuanrong. Compared to traditional Serverless function concepts, openYuanrong functions are more general-purpose, supporting dynamic creation during runtime, long-running execution, asynchronous inter-calls, statefulness, etc. They can express running instances of arbitrary distributed applications, serving a role similar to processes in a standalone OS.

Stateful Function: openYuanrong functions support statefulness, where state refers to private in-process variables that can be accessed and modified during program execution. Typical states include static variables within a process, member variables in object-oriented programming, etc. Based on the stateful function concept, openYuanrong provides multi-language programming interfaces, allowing classes originally developed using Python, Java, and C++ to be automatically converted into openYuanrong stateful functions for execution.

Stateless Function: Stateless functions are a special case of stateful functions where execution does not depend on state, only on input parameters. Based on the stateless function concept, openYuanrong provides multi-language programming interfaces, allowing functions originally developed using Python, Java, and C++ to be automatically converted into openYuanrong stateless functions for execution.

Data Object: Data objects are in-memory data that can be shared and accessed across nodes between multiple openYuanrong functions, supporting high-performance put/get access and modification based on shared memory; additionally, they can serve as function call parameters and return values, automatically distributed and shared, and support asynchronous Future semantics: for example, calling a function can return a Data Object Future reference, at which point the object reference can also be used as a new function call parameter; openYuanrong automatically resolves references during runtime and manages data object lifecycles through automatic distributed reference counting.

Data Stream: Data streams are ordered unbounded in-memory datasets that can be passed and shared across nodes between multiple openYuanrong functions, supporting high-performance pub/sub access based on shared memory, and supporting one-to-one, one-to-many, many-to-one, and other publish/subscribe patterns. Data streams facilitate decoupling of multiple different functions, enabling asynchronous streaming data transfer and computation between functions.

Master Node: A node in the openYuanrong cluster that contains both control plane (cluster management, scheduling, etc.) and data plane (running distributed tasks) components. Control plane components include function master, and data plane components include function proxy, function agent, runtime manager, and data worker.

Agent Node (Worker Node): A node in the openYuanrong cluster that contains only data plane (running distributed tasks) components, including function proxy, function agent, runtime manager, and data worker. Multiple agent nodes can be deployed on a single host.

Function System: One of the openYuanrong systems, providing large-scale distributed dynamic scheduling, supporting rapid elastic scaling and cross-node migration of function instances, achieving efficient cluster resource utilization.

Data System: One of the openYuanrong systems, providing heterogeneous distributed multi-level caching, supporting Object and Stream semantics, achieving high-performance data sharing and transfer between function instances.

Driver: The name of an application’s startup process. For example, Python startup scripts, C++ binary executable files, or Java executable jar packages.

Host Deployment (Process Deployment): Starting all openYuanrong components as processes on a host, providing basic health monitoring and process re-launch mechanisms. Process deployment is typically used for lightweight local validation and other scenarios with lower requirements for availability, reliability, and isolation.