Disappointingly small number of deployments, many open source software and open API organizations (ONF, Linux Foundation, MEF, TM Forum, OCP, etc), pop-up consortiums (Cloud RAN, Open RAN, other disaggregated hardware), defunct standards organizations (e.g. ETSI, ITU-T, IEEE) that only produce functional requirements, reference architectures, and white papers or none of the above. Nothing that can be actually implemented via standardized exposed interfaces or APIs.
Tuesday April 30th and Wednesday May 1st I spent the entire day and early evening at the Layer 1,2,3 Network Transformation Congress which assessed the state of SDN, NFV, Open Source MANO (OSM), Open APIs (TM Forum and MEF), other Open Source management software, and topics related to what network operators have been talking about for at least eight years- computer controlled network automation and orchestration of services (sometimes referred to as service chaining). Contrary to the rah, rah cheerleader talk from a few network operators (especially AT&T), telco deployment of this new age open source software for automation and control of networks has been very slow. NFV actual deployments are minimal (if not zero) and SDN has become a marketing term that can mean any software control of network functions. Every network operator and cloud service provider uses different protocols, many of which they invented (e.g. Google’s routing protocol for DCI) along with a sprinkling of open source code (such as a SDN Controller).
Decades of man years has been invested in network operator proprietary network management software, which is used to provision new services, keep track and maintain existing services, facilitate moves and changes. One speaker said that he’d like to see light touch provisioning rather than zero touch. Another said that they stack the new automation, provisioning and orchestration software on top of their legacy software
For the cloud giants (e.g. Amazon, Google, Tencent, etc), it has been done, but in almost a totally proprietary fashion with almost all the network automation, control and management done using in house generated code. Amazon spoke at the conference and, in response to this author’s question, suggested the different types of network access for AWS. Microsoft spoke, not about Azure but their private enterprise network which doesn’t use any open source code. Moreover, it took two years to get 22 new sites connected via direct internet connections (<600M bit/sec) that would normally be served by copper lines (bonded DSL or short reach fiber).
Selected Quotes from Conference Participants:
Long time colleague Craig Matsumoto (whom I met when he was EE Times, but now at 451 Research) coined a new term during his presentation – “software programmable interconnection” (SPI) for data centers. Craig said: “We talk a lot about telcos. The question is what does network transformation mean for the data center world? What are they doing about it? We came up with this new term, software programmable interconnection (SPI) . It’s basically about the idea that data centers connect with one another with a fabric.” In this author’s opinion the SPI term captures the wide variety of software being used within and between data centers!
“For me covering data centers after covering telcos for so long, they’ve (data center operators) talked to me about using the SDN for pretty much anything that involved automation and the network. Anything that has software is SDN to them. We came up with a different term as a good way to encapsulate that some kind of software is being used that might or might not be SDN,” Matsumoto added.
Tuesday’s keynote speaker and Wednesday moderator Roy Chua, Founder and Principal of AvidThink – a boutique market research firm:
“With regard to the key takeaways, I think you’ve captured them. I was very impressed at the level of candor in the discussions and presentations. I liked the concrete examples and quantification of NFV uptake challenges and the recognition that we need to solve constrained problems than try to boil the ocean. There was definitely good content…..Appreciate all the excellent questions and enjoyed the discussion at lunch. And I am most grateful for your endorsement of the analysis that I do.”
This author recommends only a select few (<5) networking market analysts that do primary market research. Roy is one of those select few!
From Kaustubha Parkhi, Principal Analyst at Insight Research (a well respected Indian market research firm):
“There is no doubt that LSO  is essential. Equally essential is the pruning of its objectives and scope, which becomes a bit overwhelming at times. The objectives, in the present form are so broad-based that they cover everything from billing functions to network equipment deployment.” –>More on LSO in a forthcoming IEEE Techblog article.
Note 2. LSO (Lifecycle Service Orchestration) is the set of MEF-defined specifications enabling standardized service orchestration based on standardized lifecycles of end-to-end connectivity services across one or more network service domains. A key contribution is open APIs – to automate the entire lifecycle for services orchestrated across multiple provider networks and multiple technology domains within a provider network. LSO enables service providers to transition from a silo-structured BSS/OSS approach towards flexible end-to-end orchestration that unleashes the value of SDN and NFV. Standardized LSO APIs are critical for enabling agile, assured, and orchestrated services over automated, virtualized, and interconnected networks worldwide.
Above illustration courtesy of MEF
I was pleasantly surprised by the honesty (if not brutal frankness) of the speakers. What a refreshing change from the never ending hype, exaggeration and lies one hears at most networking conferences – including the IEEE 5G Summits :-((.
With over 20 pages of handwritten notes and so many important things revealed, I am not able to write a detailed conference summary report on this free website. Hence, I solicit readers to email me what they’d like me to cover in future posts, after reading the conference agenda for Tuesday- Day 1 and Wednesday -Day 2.
Please remember that the IEEE Techblog does not accept advertisements so we can tell the real truth. Also we don’t charge for viewing posts or comments (no pay wall). Finally, this author has managed and contributed to this and predecessor website (community.comsoc.org) for over 10 years without any pay.
You may contact this author at: [email protected]