Just like in Gene Roddenberry’s “Star Trek,” where the Borg subsumed all other races with the tag line “resistance is futile,” programmable fabrics are fast becoming the inevitable way of the future; and whether we consume them underneath a more traditional full stack router or more directly using modern day programmable fabric controllers, they are set to become a part of the networking fabric.
Programmable fabrics have been predicted to revolutionize the network space for quite some time; however, we’re now seeing several key drivers that point to this technology becoming a reality sooner rather than later.
There are four major use cases:
1. Next Generation Network Fabric: providing a traffic engineered, multi-tenanted and scalable network fabric that doesn’t suffer from the problems with today’s mixed overlay/underlay fabrics.
2. 5G and Edge: converging and offloading fixed and wireless services at the edge and in the process reducing latency, jitter, power and space and freeing up server cores for new edge applications.
3. Telemetry & Autonomous networks: using in-band telemetry to drive more consistent and intelligent forwarding decisions in the fabric.
4. Automation: enabling CI/CD pipelines to roll network changes into production more quickly, reliably and efficiently.
These drivers require a different architectural approach for the network fabric than has been required in the past. The traditional router/switch architecture running a full stack of routing protocols on each network device will be instead moved more centrally to run atop a fabric controller for that domain. They will still be interacting with other existing domains using traditional routing protocols, but allowing the above use cases within their programmable fabric domain as shown in the following diagram (note: the NG-SDN Domain is the programmable fabric):
While these programmable fabric domains may initially be introduced into smaller areas of the telecommunication provider’s network like the access or edge networks, over time they’ll likely be joined, and we may see a hierarchy of controllers used to control the end to end network similar to Google’s global SDN networks.
What is a “Programmable Fabric”?
A “Programmable Fabric” is the next generation of the Software Defined Network (i.e. ONF’s NG-SDN) in that it allows programmability all the way down to the forwarding pipeline. This programmability can then be used by network functions, allowing them to be fully disaggregated and virtualized and truly defined in software to deploy where and as needed in the network. Programmable Fabrics are the inevitable next step in the next generation of networks.
Dell Technologies is helping our Telecommunication Service Providers in the following ways:
◉ More cost efficient: Use of merchant silicon with strategic vendor support will help manage costs while scaling the network.
◉ More Open: An open and standards-based solution will allow the service provider to take advantage of industry innovations more easily while ensuring no vendor lock-in and will be key to driving down costs in the network.
◉ Multi-tenanted: The ability to define multiple different virtual networks in software on the same physical network for the purposes of multiple tenants or applications while keeping all the other characteristics required by a service provider (i.e. traffic engineering and telemetry).
◉ Traffic engineered: The ability to allow tenants or applications to signal priority and resourcing to the network and have the network provide that service (i.e. bandwidth, latency, jitter, specified path, redundancy characteristics).
◉ Telemetry: In-band telemetry that allows real time recording of packet metrics (i.e. latency, jitter, drops) on a per session basis is key to the operational support of the solution.
◉ Intelligent Control Loop Feedback: Machine learning can be applied to the telemetry being received from the network to provide intelligent control loop feedback to prevent or repair network failures and ensure the network is being operated correctly and efficiently.
◉ Continuous Integration and Continuous Deployment (CI/CD): Will be key to allowing service providers to integrate changes more easily and quickly into the network, allowing full operational readiness tests to be carried out before changes are deployed into a production network.
◉ Network Function offload: The ability to offload the user plane portion of network functions into the programmable fabric (BNG for fixed or UPF for wireless subscriber termination are examples of network functions that could be offloaded into the programmable fabric).
Programmable fabrics have been predicted to revolutionize the network space for quite some time; however, we’re now seeing several key drivers that point to this technology becoming a reality sooner rather than later.
There are four major use cases:
1. Next Generation Network Fabric: providing a traffic engineered, multi-tenanted and scalable network fabric that doesn’t suffer from the problems with today’s mixed overlay/underlay fabrics.
2. 5G and Edge: converging and offloading fixed and wireless services at the edge and in the process reducing latency, jitter, power and space and freeing up server cores for new edge applications.
3. Telemetry & Autonomous networks: using in-band telemetry to drive more consistent and intelligent forwarding decisions in the fabric.
4. Automation: enabling CI/CD pipelines to roll network changes into production more quickly, reliably and efficiently.
These drivers require a different architectural approach for the network fabric than has been required in the past. The traditional router/switch architecture running a full stack of routing protocols on each network device will be instead moved more centrally to run atop a fabric controller for that domain. They will still be interacting with other existing domains using traditional routing protocols, but allowing the above use cases within their programmable fabric domain as shown in the following diagram (note: the NG-SDN Domain is the programmable fabric):
What is a “Programmable Fabric”?
A “Programmable Fabric” is the next generation of the Software Defined Network (i.e. ONF’s NG-SDN) in that it allows programmability all the way down to the forwarding pipeline. This programmability can then be used by network functions, allowing them to be fully disaggregated and virtualized and truly defined in software to deploy where and as needed in the network. Programmable Fabrics are the inevitable next step in the next generation of networks.
Dell Technologies is helping our Telecommunication Service Providers in the following ways:
◉ More cost efficient: Use of merchant silicon with strategic vendor support will help manage costs while scaling the network.
◉ More Open: An open and standards-based solution will allow the service provider to take advantage of industry innovations more easily while ensuring no vendor lock-in and will be key to driving down costs in the network.
◉ Multi-tenanted: The ability to define multiple different virtual networks in software on the same physical network for the purposes of multiple tenants or applications while keeping all the other characteristics required by a service provider (i.e. traffic engineering and telemetry).
◉ Traffic engineered: The ability to allow tenants or applications to signal priority and resourcing to the network and have the network provide that service (i.e. bandwidth, latency, jitter, specified path, redundancy characteristics).
◉ Telemetry: In-band telemetry that allows real time recording of packet metrics (i.e. latency, jitter, drops) on a per session basis is key to the operational support of the solution.
◉ Intelligent Control Loop Feedback: Machine learning can be applied to the telemetry being received from the network to provide intelligent control loop feedback to prevent or repair network failures and ensure the network is being operated correctly and efficiently.
◉ Continuous Integration and Continuous Deployment (CI/CD): Will be key to allowing service providers to integrate changes more easily and quickly into the network, allowing full operational readiness tests to be carried out before changes are deployed into a production network.
◉ Network Function offload: The ability to offload the user plane portion of network functions into the programmable fabric (BNG for fixed or UPF for wireless subscriber termination are examples of network functions that could be offloaded into the programmable fabric).
Source: dellemc.com
Posts
0 comments:
Post a Comment