SDN architectures

ESDN: OSS Implosion

NEW YORK -- Ethernet & SDN Expo -- One persistent message coming from this event is that today's operations and support systems (OSSs) are thoroughly unprepared to address the requirements of the virtualized networks that service providers are planning to deploy.

This view was repeatedly shared, often in bleak terms, in multiple Wednesday panels and in a Heavy Reading analyst wrapup Thursday morning.

  • "The OSS needs new clothes." -- Nirav Modi, director of software innovations, Cyan Inc.
  • "We need to re-evaluate the whole architecture, radically." -- Margaret Chiosi, distinguished network architect, AT&T Inc. (NYSE: T)
  • "The OSS will have to evolve significantly." -- Manish Gulyani, vice president of product marketing, Alcatel-Lucent (NYSE: ALU)
  • "The current OSS is broken and must be replaced. It's been patched for the last 20 years." -- Ralph Santitoro, director of strategic market development, Fujitsu Network Communications Inc.
  • "SDN/NFV will require a complete transformation of the OSS and tremendous re-skilling by the operators. It will also mean significant redundancies... whole groups and departments may go away." -- Graham Finnie, chief analyst, Heavy Reading

The opinions that were offered were complex but not always detailed. Chiosi talked about three critical areas that have to be reformulated: orchestration for datacenters, network management for network functions virtualization (NFV), and SDN controllers. She said AT&T needs self-provisioning capabilities for its customers and self-optimizing networks (SON).

And though she stressed open application programming interfaces in her earlier presentation, Chiosi said standardized interfaces are hardly enough to address the OSS issues. Instead, the entire OSS infrastructure needs an overhaul -- a commonly held opinion that leads on to wonder whether this is a job for software engineers or demolition experts.

Several speakers discussed the possibility of creating layers of abstraction that essentially surround the legacy OSSs and pull data from them, as needed, without using them as management tools. Prayson Pate, chief technologist at Overture Networks Inc. , suggested creating abstraction at different layers, so that necessary controls are housed within each layer but aren't shared among other layers, which might create unnecessary complexity.

The other commonly discussed approach seems a lot like a cap-and-grow strategy. It involves investing in more flexible technology to manage newer services and virtualized infrastructure while allowing the legacy OSSs to continue to handle legacy services -- perhaps in the belief that they'll quietly fade away.

That doesn't mean there wasn't some specific advice. Heavy Reading senior analyst Caroline Chappell said that carriers often cite the need for OSS updates as an excuse for not shortening their often months-long provisioning cycles. "The tools exist today to do this in minutes. I've seen it done," she said. "There's no reason why service providers can't be [turning up] services much faster than they do today."

But much of the discussion around preparing OSSs for the future virtualized network is still theoretical, which is what made for very lively debate here at the ESDN Expo.

— Carol Wilson, Editor-at-Large, Light Reading

Page 1 / 2   >   >>
mendyk 10/9/2013 | 3:03:38 PM
Re: Why not New PLUS Old? Since we only THINK we know what the future will bring, it's almost certain that we will be in a "work in progress" mode for a long, long time. That's not necessarily a bad thing (it keeps people employed, for instance). The message about real-world telecom does get distorted somewhat because the people who command the microphone usually are pushing a particular point of view or agenda. That's kind of the essence of the hype cycle.
shawn_mc 10/9/2013 | 2:48:38 PM
Re: Why not New PLUS Old? Isn't OSS change as inevitable as network change?  We'd all like some (more?) future-proofing built into the OSS to make it as accomodating of new network technologies as possible, but perhaps that's where network abstraction comes into play. Or perhaps some form of OSS transformation & evolution is always going to be required as the network continues to evolve? 
mendyk 10/9/2013 | 2:33:31 PM
Re: Why not New PLUS Old? This is a very important point: Change happens gradually, whether we like it or not. Regarding OSS, though, the frustration is understandable, given that OSS transformation is well into its second decade with no clear end in sight.
shawn_mc 10/9/2013 | 2:25:00 PM
Why not New PLUS Old? Every time I read articles about the latest technology to take the network by storm, I'm always alarmed that the mentality seems to be cap X (or throw out X) and start deploying Y. There are not too many greenfield networks being built these days and any new technology needs a migration or evolution strategy where the tools and OSS systems can work with the old network AND the new network. Such platofrms do exist today, and are required to make X+Y work together during the evolution of the network...
DOShea 10/7/2013 | 8:54:51 PM
Another OSS implosion I'm having deja vu--didn;t the industry already blow up its OSSs a couple of times in the last 20 years? If those were just patch-up jobs, we sure talked about them like they were much more.
sam masud 10/7/2013 | 3:54:49 PM
Re: So being a gadfly I think we are 5-10 years away from SDN deployment--regardless of whether existing OSS have to be scrapped or not.
abanerjee 10/5/2013 | 9:02:46 AM
OSS complexity will increase OSS will be the Achilles heel of SDN. There are many reasons for that and many of them I have highlighted in my previous blog. http://www.lightreading.com/author.asp?section_id=193&doc_id=702157

We have seen OSS transformation projects fail in the past and we all know Tier 1 operators spends millions on maintaining their legacy OSS platforms. Are they going to replace them or rewrite them just to make SDN work? Ii doubt that as it will cost in either case millions of dollars. My take as an analyst conducting OSS research for many years this will create another OSS silo as and when SDN becomes mainstream and cause OSS complication to increase.

Liz Greenberg 10/4/2013 | 3:45:01 PM
People, training, and time I think that it is easy to underestimate the amount of time required to train everybody in an OSS so that they actually understand what they are looking at and what failures mean, how to work around them etc. The "former/current" network with switches, tandems, etc. are being replaced with extremely complex abstract  computers that are going to take a while for OSS employees to learn well.  Overlay is a great way to go as some employees can work each and as more of the network transitions, more employees can as well.
brookseven 10/4/2013 | 12:41:24 PM
Re: So being a gadfly Carol,

I think that is always the challenge with a new technology.  In one of my past employers, the plan was to build a platform that would require a swap out of an entire level of network element before value can be added.

I think this is possibly a challenge here, especially if you have to interface to the older environment.  It means that there are probably significant limitations in what the new technology will be able to accomplish in a mixed environment.


Carol Wilson 10/4/2013 | 11:43:33 AM
Re: So being a gadfly That was one of the points of discussion - whether blowing up the legacy OSS isn't going to slow down the deployment of SDN/NFV -- which is why some folks are talking more cap and grow. 
Page 1 / 2   >   >>
Sign In