NFV

Expand all | Collapse all

In production services built from NFV/SDN

  • 1.  In production services built from NFV/SDN

    Posted Oct 02, 2017 11:41
    I have been hearing about SDN/NFV trials for more than 5 years. All of the trials are very "successful" and show all kinds of improvements in time to deploy services, ability to scale up and scale down on demand, improvements in operating expenses, greater flexibility and agility you name it. So ... If the Business Case for SDN/NFV is so compelling, why is it taking so long?

    I still see very few IN PRODUCTION services built from SDN/NFV systems and platforms. Does anyone have any thoughts on this?

    ------------------------------
    Craig Farrell
    IBM Corporation
    ------------------------------


  • 2.  RE: In production services built from NFV/SDN

    Posted Oct 03, 2017 02:30
    We actually did some research on this and some of the biggest challenges relate to the commercial relationship between supplier and operator, while other issues are similar to the inhibitors to digital transformation more broadly – culture, skills and organizational alignment.

    The lack of a compelling business case was the biggest business inhibitor cited by both CSPs and suppliers, with 42 percent of CSPs and 43 percent of suppliers saying it is a 'very serious' or 'moderately serious' challenge.

    Read page 10 in our recently released Digital Transformation Tracker for more detailed info.



    ------------------------------
    Arti Mehta
    Editor, InformTM Forum
    ------------------------------



  • 3.  RE: In production services built from NFV/SDN

    TM Forum Member
    Posted Oct 03, 2017 10:31
    ​Craig,
    There does seem to be some disquiet around NFV. If you look at Tom Nolle's post he has a interesting viewpoint especially in

    Would Savings from NFV or Lifecycle Automation Fund Innovation?


    HIs conclusion is that the CAPEX and especially OPEX saving are not properly established outside some very narrow missions including vCPE . He also has issues with the ETIS NFV Automation approach described in other poste

    Some of the suggestions he makes around intent based management are already addressed in the ZOOM HIP work  in

    TR262 Hybrid Infrastructure Platform Blueprint R17.0.0



    ------------------------------
    Dave Milham
    TM Forum Chief Architect
    ------------------------------



  • 4.  RE: In production services built from NFV/SDN

    Posted Oct 04, 2017 02:46
    I must admit I find this article quite compelling. I also had come to the conclusion that there was a problem with the business case for NFV which was why operators were taking so long to adopt it. I talked about the problem with the business case in my articles published last year.   https://inform.tmforum.org/features-and-analysis/2016/11/dilemma-implementing-nfv-services-taking-long/
     
    The thing I liked about this other article is that it puts some estimates of the numbers involved which highlights the size of the SDN/NFV adoption business case problem. 
     
    Regards,
    Craig.
     
     
    Craig Farrell.
    Chief Technology Officer, Global Telecom Industry,
    Email: craig.farrell@us.ibm.com
    Cell: 858 342 0433






  • 5.  RE: In production services built from NFV/SDN

    Posted Oct 04, 2017 03:44
    I would generally disagree with this, there is a fairly wide number of NFV deployments (with SDN based overlay infrastructure) in production, especially for control / management plane applications (IMS, PCRF etc.) and more CSPs are starting to break ground on user-plane intensive workloads (vEPC, vCPE, SD-WAN etc).

    However, there are potentially two things that need to be fully understood by the CSP for a successful NFV roadmap:

    1. This is NOT about a CAPEX or OPEX saving in the short-term, there will be changes in the long term that will bring an overall reduction in TCO, but in my opinion deploying one or two VNFs doesn't necessarily save you tonnes of money - but building the right platform does give you the agility to move faster than in a physical world

    2. To do this successfully requires some organizational changes that not everyone is ready for, the outsourcing to NEP model is not (IMO) the best approach to deliver a valuable NFV stack

    In addition to these two challenges, the market is not significantly mature enough with regards to VNF Managers (VNFMs) and Orchestration platforms (MANO) - these are getting better and there are some clear though leaders in the MANO space - so this is ongoing.

    The organizational buy-in from top down is also very important, there needs to be a solid commitment across the organization to ensure the venture is successful.

    ------------------------------
    Gary

    ------------------------------



  • 6.  RE: In production services built from NFV/SDN

    Posted Oct 11, 2017 07:23
    Hi Gary,
      Since you disagree and point out there are many "in production" SDN/NFV deployments, can you tell me the ROI for any of these deployments? Can you find me some actual ROI numbers please?  Not estimates, not "expected" improvements, some actual measured numbers?

    regards,
    craig.

    ------------------------------
    Craig Farrell
    IBM Corporation
    ------------------------------



  • 7.  RE: In production services built from NFV/SDN

    Posted Oct 12, 2017 04:15
    Craig,

    Unfortunately I cannot, I can only say for certainty that there are production NFV/SDN deployments as I've designed some myself and actually worked with multiple customers who have live NFV Production environments today.

    That being said, the initial RIO I guess is a little too early to measure for some of these customers, I guess the only people actually holding actual numbers in terms of ROI, Metrics and KPIs would be the CSPs.


    ------------------------------
    Gary Day
    ------------------------------



  • 8.  RE: In production services built from NFV/SDN

    Posted Oct 13, 2017 02:04
    Hi Gary,

    We also believe there is a space for NFV. But too many people position it as Network Automation and Orchestration. And its not. NFV is substituting discrete hardware with its own pro's and con's as discrete hardware has its own pro's and con's as well.

    ------------------------------
    Wim Horseling
    NetYCE
    ------------------------------



  • 9.  RE: In production services built from NFV/SDN

    TM Forum Member
    Posted Oct 04, 2017 03:58
    You can debate the micro issues as much as you want but the macro issue is unassailable.
    The old adage goes "Never bet against the Internet!".

    NFV is currently in the "trough of disillusionment" phase of the technology life cycle.
    Hold on, it get's much, much better!

    ------------------------------
    Vance Shipley
    SigScale Global Inc.
    ------------------------------



  • 10.  RE: In production services built from NFV/SDN

    Posted Oct 05, 2017 01:59
    Hi
    Thanks for good discussion.

    It is not fair to pinpoint on one or two issues while looking at NFV adoption. Instead it needs a comprehensive look for complete business ecosystem that should cover customer, products, regulatory environment and must be part of CSP strategy. At technology front as well more evolution and standardization need to happen so its a journey that is still away from end point.

    Realistically its much slower than anticipated before couple of years. If we go back to google some of projections and targets or visions published at that time and see the reality as of today then there is wide gap. And my view is that it'll not change in next 12-18 months either. We'll see success in deployments with NFV remaining lightweight, excluding pureplay LTE core (including IMS) and CPE.

    Out of these CPE ( including edge routers) because it was always major operational headache and involved truck-rolls for most of faults i.e. cost and difficult to attend both. Moreover it is not an issue towards rest of network. So best target.
    IMS and LTE core, because these are relatively new, lightweight and on flat IP design. So a lot of scope for innovation & collaboration.

    But I am not dreaming for big bang transformation from physical to virtual for incumbents in next 1-2 years, neither I see any business compulsion when CSPs have bigger transformation to do in operations, customer engagement, system rationalization etc to be better prepared for NFV leap.
    Hence its journey towards maturity from both CSP and technology evolution perspective that will converge after sometime.


    ------------------------------
    Dharmendra Misra
    IBM Corporation
    ------------------------------



  • 11.  RE: In production services built from NFV/SDN

    Posted Oct 12, 2017 03:52
    We believe NFV/SDN only isn't complete. 
    The biggest cost in a 5 year life cycle of a network is not the initial procurement nor its initial set-up. Its what comes after that. Service (assurance) monitoring, and executing the change requests. Moves-Adds-Changes or CRUD's Create-Read-Update-Delete.  To NFV deliver ROI this is what will need to be addressed and become an process without humans involved. This is what shareholders want to see before putting big money into NFV/SDN. Having a VNF like a router spinned-up still requires someone to configure it. By a human. The service the customer gets (VPN, VLAN, MPLS, Firewall etc) all boils down how the device is configured -what you want to sell as provider in the competitive landscape - your design/services catalogue. As far as we see it, equal to traditional networks the 100% truth of how the devices are really configured lives in the network device itself. Alternatively in an vendor specific element manager. As we see it in NFV its still element managers - SILO based solutions that configure functions. Which don't bring the end-to-end service for the customer as networks will remain multivendor and hybrid, mix of PNF and VNF. Having a repository of configuration files by itself is not good enough.

    You can only automate and orchestrate - configure a new circuit to accommodate expected traffic if this is designed upfront. As what intent based in shooting for.The configuration file needs to be a deliverable of collaborative (cross teams and individuals) digitised design and deployment process (top down). This way your designs and services (what you sell) are converted into data that can be manipulated. An API call from an orchestrator, Service assurance system or SIEM can reconfigure the network. It robotised amends a configuration file or loads a new one. to shut down a port, increase bandwidth, create circuits in the right sequence on multi-domain nodes ets. So as a note its also required to have a process in place that secures design-build-operate in such way that also the SLA's between the processes are digitised. Which makes it somewhat hard to deploy such a solution as it means you impact the organisation.

    Such a solution can co-exist with NFV/SDN, actually it would complement it.

    ------------------------------
    Wim Horseling
    NetYCE
    ------------------------------