Skip to end of metadata
Go to start of metadata

Sukhdev Kapur


Project Technical Lead:               Sukhdev Kapur (sukhdev@juniper.net)


Project Committers detail:

Committer

Committer

Company

Committer

Contact Info

Committer Bio Committer Picture 

Self Nominate for PTL (Y/N)







Sukhdedv Kapur

Juniper



Y
Gleb GalkinJuniper


Alexandre LevineJuniper
alexandrelevine@gmail.com>



Andrey PavlovJuniper


Paul CarverAT&T


Oleg BravoJuniper


  • No labels

2 Comments


  1. Network Cloud Integration with Tungsten Fabric

    Attributes

    Description

    Informational

    Type

    New

     New submission for integrating Tungsten Fabric with Network Cloud Blueprint.

    Industry Sector

    Telco, Edge Cloud and 5G

    The blueprint empowers Network Cloud architecture to use single SDN Controller (Tungsten Fabric) for all type of Containers, Virtual Machines, and Bare Metal Servers.

    Telco Cloud, Edge Cloud, and 5G industry sector.

    Business driver

    Empower Edge sites with basic and advance networking features via single SDN controller.
    Distributed Compute Architecture for Edge site.
    To secure different types of workloads running at the edge with a single SDN controller.

    Today most of the vendors' telco applications are virtualized as VNFs or still using legacy PNF.  The vendors are already working on containerization of their apps. Data Center infrastructure supporting all different type of workloads (PNF, VNF, CNF) is required to support fully distributed networking stack, not only on Core site but at the Edge sites as well. Tungsten Fabric as a fully distributed networking SDN controller is an ideal fit for providing support for all type of workload by delivering e2e provisioning, management and monitoring of overlay and underlay networking. 

    SDN controller for 5G edge use-cases to provide primary and advance networking features.

    Business use cases

    Use-Cases:

    • Distributed Edge computing using TF distributed compute (Remote Compute) architecture
    • Service Chaining at the Edge sites
    • Unified SDN controller for VNFs, CNFs
    • Fabric provisioning for SR-IOV workload
    • Edge workload security
    • TF support for Helm and Ansible base deployments


    Business Cost - Initial Build Cost Target Objective

    Tungsten Fabric central site requires 9 VMs ( 3 Control, 3 Analytics & 3 Analytics-DB) for full HA (2n+1) where n=1 along with TF vRouter running in compute at the edge site.

    A centralized TF Cluster on an on-prem or public cloud can support multiple edge sites using "Sub-Cluster-Id" feature.   One can keep creating sub-clusters for different edge sites and keep expanding Edge infrastructure at scale. 

    Business Cost – Target Operational Objective

    1. TF Edge Cloud Controller deployable at Central offices with 9 servers (VMs or BMS) should incur low operational costs per year
    2. TF support ISSU (In Service Software Upgrade) upgrade of the Edge cloud without impacting the availability of the edge applications
    3. TF support role based access controls, Single Pane of Glass control, administrative and User Based GUIs to manage all   deployments.
    4. TF supports Helm and Ansible for the automation and management tools to keep operational cost lower



    When managing multiple edge sites using Distributed Compute any networking level outage, applications failure can cause traffic loss on the distributed edge. To address such failures TF GR/LLGR (Graceful Restart & Long Live Graceful Restart) with BGP and XMPP helper protects forwarding plane on the edge sites. End-to-end monitoring and single pane of glass operation supported.

    Security need

    TF security framework for all type of workload protection and

    TF Security is fully integrated with orchestration platforms providing Security Logging Object SLO, Network Policy enforcement for microservices, Data Center security groups enforcement on fabric leaf switches and BMS "Bare Metal Server" security by running vRouter on BMS.  With TF Security, one defines policies once and automatically distribute them uniformly across all types of workload deployments. You can also use it to easily monitor and troubleshoot traffic flows inside each application and between applications. All in all, this feature will bring value towards Edge computing security framework.

    Regulations

    Tungsten Fabric is already in the production/operations of tier-1 Telco clouds. It meets all required regulations

     Today vEPC VNFs (vMME, vSGW, vPGW, vPCRF, vMMSC and many more) are running in production.

    Other restrictions

    The smallest footprint for TF is a single VM. It assumes MPLS/BGP connectivity

     This is not really a restriction, but, worth mentioning here

    Additional details

    TF can be deployed across the globe


    If the proposal includes a new Blueprint Species addition to an existing or new Blueprint Family include a completed Blueprint Species template specific to each Blueprint Species being submitted.

    Case Attributes

    Description

    Informational

    Type

    New


    Blueprint Family - Proposed   Name

    Network Cloud Integration with Tungsten Fabric


    Use Case

    Use-Cases:

    • Distributed Edge computing using TF distributed compute (Remote Compute) architecture Figure:1
    • Service Chaining at the Edge sites
    • Unified SDN controller for VNFs, CNFs
    • Fabric provisioning for SR-IOV workload
    • Edge workload security
    • TF support for Helm and Ansible base deployments


    Blueprint proposed Name

    Tungsten Fabric a unified SDN Controller


    Initial POD Cost (capex)

    The intent is to use existing Network Cloud hardware resources. This is SW upgrade solution only


    Scale & Type

    For the smallest deployment, this requires 9 VMs on central site and 1 VM on remote site. For large deployments, this could span to large number of bare metal machines (computes)


    Applications

    5G Core and Edge Sites.


    Power Restrictions

    N/A


    Infrastructure orchestration

    This is integrated solution with existing Network Cloud, therefore, it uses the same orchestrators and infrastructure as NC


    SDN

    Tungsten Fabric Kernel vRouter, DPDK vRouter, SR-IOV and SmartNIC


    Workload Type

    Any workload: Virtual Machines, BareMetal Servers and Containers

    VNF, CNF and BMS


    Additional Details

    Tungsten Fabric is fully distributed networking stack from day one and now with new microservices architecture, design and the Edge Cloud features set, it is now ready for Edge computing distributed networking framework. TF SDN Controller provides seamless and full integration between different types of workloads VNFs, CNFs and PNFs using single Networking stack used by different orchestration platforms like OpenStack, Kubernetes, and, Fabric & BMS manager.  TF SDN Controller in this mode of operation will work as single entity running in on-prem (core, distributed core or edge site) or public cloud (AWS, Azure, GCP or Packet.net) and fully integrated with OpenStack Neutron Plugin, Kubernetes CNI, and BMS/Fabric management for all type of Edge computing workloads. As captured in the figure:1 TF SDN Controller integrated with OpenStack via Neutron, with Kubernetes as a CNI and with Fabric/BMS Manager using TF Fabric Manager function. All TF basic and advance networking features set highlighted in figure:1 will be available for different Edge computing use-cases.