What is Edge Computing?
Unlike the core infrastructure that is generally uniform and centralized, edge infrastructure consists of many points of presence with architecture varying as a function of proximity to the end user and the type of application.
Deploying and running cost efficient edge infrastructure forces service providers to rethink infrastructure requirements across the following two categories:
Baseline requirements generic to all edge points-of-presence, driven by the highly distributed nature of edge infrastructure
Footprint architecture requirements specific to a particular point-of-presence, driven by the proximity to the end user and the type of application
Edge Infrastructure Baseline Requirements
MCP Edge satisfies baseline requirements common to all deployments of edge cloud infrastructure.
To keep operational costs down, CSPs need a single centralized management across multiple edge deployments.
MCP Edge is capable of managing the lifecycle of edge locations via a centralized management plane powered by DriveTrain .
Edge/NFV orchestration software, VNF managers, and 3rd party applications require industry standard northbound APIs for interoperability.
MCP Edge uses an open standards first approach. Based on OpenStack and Kubernetes, MCP Edge northbound APIs are 100% compatible with industry requirements.
Access-network edge locations connect to aggregation edge, which in turn connect to the regional POP. To get dynamic end-to-end services, this connectivity and related configuration need to be fully automated.
MCP Edge uses BGP VPN to connect various edge locations. The connectivity and configuration can be automated through OpenDaylight or proprietary SDN controller software from Mirantis partners .
An application or project must be able to span multiple edge locations for reasons such as mobility, subscriber correlation, and others. This requires application identity to work across edge locations.
MCP Edge uses a federated OpenStack Keystone catalog to provide identity federation across multiple edge locations. In future, this will also be extended to k8s.
Application Driven Footprint Requirements
Applications running on the edge drive the need for different reference architectures. MCP DriveTrain enables this flexibility both upon initial deployment and subsequent tuning.
App Driven Footprint Requirements
High bandwidth applications that require very high downstream bandwidth from the edge to the end user, e.g. multiple 4K stream
NFV applications that require VNFs to be hosted on the edge, e.g. vRAN, vOLT, vCPE, SD-WAN
Analytics applications that are compute/GPU heavy, e.g. AI/ML, Big Data
Low latency applications that have to respond to a user request within a few milliseconds, e.g. AR/VR, tactile internet
MCP Edge Features
MCP Edge provides a flexible reference architecture both for initial deployment and post-deployment changes. Variations such as these are accommodated:
- VM vs Container
- Specific software services
- Software configurations
- Dataplane acceleration features
- IA vs. ARM CPU architecture
User Proximity Driven Footprint Requirements
Proximity of the edge to the end user also drives variations in footprints and high availability requirements. MCP DriveTrain is also the mechanism to get this type of flexibility.
The access network edge requires just a few nodes with limited high availability, while a regional POP could require a multi-rack footprint with full high availability.
MCP Edge supports flexibility in footprint through different node counts, software footprint, and high availability options.