Accelerating Packet-Optical Convergence

In March, a group of leading network operators, led by Telefónica, published a white paper entitled "IP and Optical Convergence: Use Cases and Technical Requirements." The objective of this paper, as the authors write, was to outline the benefits, enablers, and challenges for IP and optical convergence, and the rationale for encouraging international collaboration to accelerate progress and development in this area.

In addition to Telefónica SA (NYSE: TEF), other contributing companies were: AXTEL Mexico; Bouygues Telecom; BT Group; China Unicom; Colt Telecom; Deutsche Telekom; KDDI; KT Corp. (Korea Telecom); Orange; and Telecom Italia.

The operator group assessed the value of a number of different IP and optical use cases. This blog focuses on multi-layer management, multi-layer planning, and multi-layer resilience, which garnered the most interest among the operator group -- and by a large margin.

The white paper looks at multi-layer resilience and focuses on an IP and optical layer resilience scheme that uses fast reroute (FRR) at the IP layer and GMPLS at the optical layer. The multi-layer network reduces operator capex by eliminating some of the requirements for 1+1 protection in the network. With traditional 1+1 protection, 50% of network capacity is reserved for backup in case a failure occurs, but a multi-layer control plane can be built to use 1:1 protection only when necessary.

While this paper does not provide multi-layer resilience savings statistics, DT research work published in 2012 demonstrated capex savings of 24-27% using a combination of IP layer and photonic layer restoration when compared to traditional network architectures implying IP layer restoration with full 1+1 protection. We have also seen equipment vendors cite even higher capex savings potential for this application.

Looking at multi-layer planning and management, a full integration of IP and optical layers can lead to better network manageability, but the complexity of full interaction (today, at least) requires a single vendor proprietary implementation. On the plus side, a proprietary integrated approach can speed up and simplify provisioning, fault management, and network planning. On the negative side, buying IP and optical layer equipment from a single vendor creates "dependencies that constrain network evolutions," according to the white paper. "Additionally, it brings a significant risk of increasing part of the OPEX, through the different upgrades and associated release qualifications since IP routers implies a high number of release updates," when compared with pure transport equipment releases.

The operators conclude that, in the case of proprietary, fully-integrated IP and optical management systems, the negatives outweigh the potential positives. The authors recommend that the industry begins working on open standards to achieve converged network management systems. This conclusion is very consistent with Heavy Reading 's findings from surveys and operator interviews during the past 18 months.

However, we continue to get pushback on this issue from various equipment vendors. A consistent counter argument we are hearing from suppliers is that, in the absence of full standards, operators will need to make proprietary decisions or lose out on the benefits of multi-layer integration completely. Granted, vendors have a strong incentive to preserve the status quo, but there is also a legitimate point being made here. With no clear cut path to standardization, how long will operators wait for multi-vendor interoperability to emerge?

— Sterling Perrin, Senior Analyst, Heavy Reading

dwx 5/23/2014 | 11:55:13 AM
Re: Restoration I think 25-30% cost savings are completely realistic in a scenario where you are just account for long-haul fiber span failures.  
sterlingperrin 5/23/2014 | 9:13:35 AM
Re: Restoration dwx,

Good points. As this project was entirely operator driven, it doesn't seem like there is a big incentive to make the applications appear more beneficial than they really are. The DT work I cite in my blog puts the capex savings at 24-27%, which is relatively modest compared to the 50-60% capex savings that I've seen cited in several vendor studies looking at the same application.

I'm not sure, but it's possible that the DT work takes into consideration the Internet traffic factors you've mentioned in your post. 


sterlingperrin 5/23/2014 | 9:04:42 AM
Full White Paper Link For anyone interested in downloading the full operator white paper (25 pages), it can be found here:



dwx 5/22/2014 | 3:23:27 PM
Restoration You have to really investigate the cost savings associated with photonic restoration.  The reality is no one uses 1+1 optical protection and builds L3 backup capacity for "Internet" IP traffic, which makes up the bulk of traffic today.   So right there you can toss out a large percentage of the cost savings because people like to tout both being implemented in networks.  The real cost savings is in not carrying as much backup L3 capacity for Internet traffic.  If you've seen slides from vendors and operators, the paradigm of "protecting" Internet traffic is going out the window and sustaining a 2-5s outage while a link goes down and comes back up is now acceptable.

By default none of this helps you in terms of equipment failure since it assumes the majority of failures are fiber cuts, which is accurate. If a L3 port goes down, you lose the capacity.  

The landscape right now is having an orchestration system/controller which can view and manage both packet and optical domains, and forcing vendors to implement open standards for programmability.  GMPLS has been going on for 15 years now, nothing has ever come from it in terms of marrying packet and optical.  

Vendors also aren't really putting packet interfaces in transport gear or OTN/DWDM interfaces in routers either.  The density and flexibility is pretty terrible for 100G in both directions.   IMHO the real opportunity is improving the interconnect between boxes with things like MacFlex so you aren't wasting either packet or optical resources.  So if I need 40G between two sites and 320G between two sites, I'm not wasting 100G and 400G to do so.  I'm using a flexible Ethernet MAC layer and then a flexible OTN container.   

Sign In