Overview

• Distributed Micro-Services System (DMSS) for Edge Networking: A Distributed System (DS) with each component or function of the DS is a light-weight Micro-Service (µ-service)

• Instances (replica) of µ-services in the DMSS can be deployed in any number in Clouds, DCs, Network Edges, Home, Metro Networks

• The DMSS has built-in µ-services for

  • Protocol-less routing and forwarding: Monolithic network functions are divided into fine-grained and light-weight Virtual Network Functions (VNF) or µ-VNFs

    • A set of Control Plane µ-VNF

    • A set of Data Plane µ-VNF: Forwarder µ-VNF

      • Forwarder µ-VNFs can run on low-cost hardware

      • Forwarder µ-VNFs can be integrated into Kubernetes (via CNI)

  • Extracting network transaction information from the overlay µ-VNF network via Deep Packet Inspection (DPI)

  • Storing network transaction information in Blockchain-like structure (DMSS Blockchain Tree)

  • Intelligently assigning Multi-core/NUMA/Multi-Cloud resources to µ-services

• The Forwarder and servers/applications are automatically composed into a µ-Service Topology (Mesh or graph in general, akin to Service Mesh)

  • Users can compose µ-Service Topology via simple configuration or via (RPC/REST) API calls

  • Multiple µ-Service Topology instances can be composed each for an application or service

    • For example, a µ-Service Topology instance for distributed IoT application or distributed Machine Learning application or healthcare IT applications

• Distributed Applications require high-availability (HA), fault-tolerance (FT), resource synchronization (RS) between multiple instances (or replica) and proper communication mechanism between instances or application components

  • • While a cluster management system like the Kubernetes (K8s) provides some level of HA/FT, it does not provide distributed application level or aware HA/FT/RS/communication mechanism, whereas DMSS provides

      • Distributed application level or aware HA/FT/RS/communication mechanism and other capabilities

      • Application knowledge-based variation of Gossip Communication and fault-tolerance (for example, one instance sees another instance faulty or down, possibly unreachable via network path, another one sees it non-faulty or up)

      • Leaderless HA/FT

      • Resource synchronization based on configurable domain or application Knowlegebase

      • µ-services communicate with each other using typical distributed application communication framework (such as RPC)

      • When multiple instances of the same µ-service scaled down or taken offline to only one, it still offers same quality of service and resources

• One of the major components of DMSS is a Service Knowledgebase Mapper (ServiceMapper) that keeps track of all µ-services, including other instances of ServiceMappers

  • A µ-service (including a newly launched ServiceMapper)

    • First discovers ServiceMappers (via discovery mechanism)

    • Discovers services it wants to interact with by hashing into the ServiceMappers Knowlegebase

    • Announces itself to all ServiceMappers

• DMSS is Implemented in Go (using RPC and other concurrency capabilities), packaged in Docker Containers as micro-services and deployed using Kubernetes.

Demos

• • DMSS µ-VNF Demo: Kubernetes (µ-VNFs run as any other pod; in Minikube and a cluster with VM on Mac and Raspberry Pis over Wifi)

  • DMSS µ-VNF Demo: Single VM (shows most of the features)

  • DMSS µ-VNF Demo: Multi-VM and VXLAN (using Docker overlay or VXLAN support natively in DMSS µ-VNF system)

Contact