Home

Distributed and decentralized MonitorAnalyzePlanExecute-Knowledge loops framework

Framework to support the development and deployment of Autonomous (Self-Adaptive) Systems

• Source Code: https://github.com/elbowz/PyMAPE
• Documentation: https://elbowz.github.io/PyMAPE - WIP

Requirements¶

• Python 3.8+
• RxPY
• PyYAML
• AIOHTTP
• FastAPI
• Uvicorn
• Redis-purse
• Influxdb-client

Key features¶

CONTAINMENT
Reuse, modularity and isolation of MAPE components as first-class entity.

COMMUNICATION INTERFACE (STANDARDIZATION)
Shared interface between components that allow stream communication, filtering, pre/post processing, data exchange communication and routing.

DISTRIBUTION
Multi-device distribution of MAPE loops and components.

DECENTRALIZED PATTERNS
Flat p2p and/or hierarchical architectures of loops and components with concerns separation. Allowing runtime pattern reconfiguration (stopping/starting, (un)linking, adding/removing).

NETWORK COMMUNICATION PARADIGMS
Different paradigms (blackboard, direct message) and protocols for various patterns interactions.

STATE / KNOWLEDGE
Distributed multi-scope (global, level, loop) Knowledge with partitioning and/or (full/partial async) replication.

Paradigms and tools¶

REACTIVE SYSTEM/PROGRAMMING AND STREAM
System reactive to external event. Pillars: Responsive, Resilient, Elastic and Message Driven. Specific case of event-driven programming to avoid callback hell.

REACTIVEX (OBSERVER PATTERN)
Observables represent a source of events. An observer subscribes to an observable to receive items emitted (Hot, Cold, Subject,etc). Pipe operators modify streams flowing through them.

ASYNCHRONOUS PROGRAMMING
Manage (I/O bound) tasks concurrency with non-blocking I/O operations.

REDIS, IN-MEMORY DATA STRUCTURE SERVER
Distributed, in-memory key-value data structure (strings, hashes, lists, (ordered) sets, queue, lock) store, cache and message broker with keyspace notifications. Partitioning and/or (full/partial async) replication.