I recently led a lively roundtable discussion on virtualization, NFV, and SDN at the recent IPX Summit in London. We had a healthy cross section of industry participants including representatives from carriers, mobile operators, equipment vendors, software centric solution vendors (like Dialogic), analysts, systems integration, and test and measurement. The question posed to the group was whether virtualization made sense for carriers and IPX operators beyond the confines of the traditional data center, and this definitely made for an interactive discussion.
The Central Office Reimagined: CORD
The title of the roundtable was misleading, I thought, because today, the central office has become or is becoming more of what we recognize now as a data center. Take the CORD initiative for example. The name alone, which is an acronym for Central Office Rearchitected as a Datacenter provides insight into the direction telecommunications is moving. The CORD project is all about virtualizing legacy hardware devices that are currently deployed in Central Offices. The reference implementation of CORD supports several elastically scalable multitenant services including access as a service.
What’s clear is that data center virtualization and automation techniques are being adopted by service providers and carriers in ways that will have a profound effect on how services are implemented and delivered.
Virtualization in the Access
There are quite a few examples of commercial offerings carriers have introduced that already incorporate SDN and NFV concepts. Some of the examples discussed at the roundtable included services and announcements such as:
- DTAG’s VPN service that includes firewall capabilities, web security, site to site connectivity, and customer self-serve features like subscription, activation, and modification.
- SD WAN or vWAN services involving SDN and NFV technology to deliver high performance, encrypted, traffic optimized bandwidth on demand across standard IP pipes
- PCCW Global’s VPN and Cloud offerings which feature on demand, self-service connectivity
- Telstra’s PEN SDN-based service that allows customers to connect their network services globally and with those of other PEN customers without having to wait for the delivery and installation of hardware appliances.
While the on-demand and self service capabilities of these services are appealing, there is an underlying physical resource aspect that can’t be overlooked. One of the operators indicated that these services were being primarily used to provide connectivity to Amazon and Azure. And while they definitely provide more traction to carriers and enable subscribers to adjust bandwidth as needed or burst when required, there is still a limitation in the access of the physical pipe used to service the customers. But the consensus was that virtualization in the access allows for:
- Rapid service turn-up
- A migration away from spreadsheet provisioning
- A high degree of automation and programmability
Regulations and Virtualization
One surprising insight from the roundtable was the impact that regulations played when it came to virtualization. This could become a particularly sticky issue in Europe where there are “some of the strictest data privacy regulations in the world.” According to those at the roundtable, regulation is an issue that could make certain uses of public cloud and virtualization not feasible or limit the services supported to the confines of the country served. This could impact operators providing hosted services via cloud infrastructure.
Skills Gap Still an Issue
Another roadblock on the road to virtualization observed by a systems integrator is the skillset gap within the operators when it comes to virtualization and cloud technology. Virtualization, NFV, and SDN introduce an entirely new set of terms, tools, and administrative practices that in many ways are orthogonal to traditional network and telecom deployments. Experience in hypervisors and cloud management systems like OpenStack require an expanded set of skills and scripting expertise with languages like YAML and YANG, and also requires an understanding of how the various OpenStack modules like Nova, Neutron, Glance and Ceilometer come into play. I touched on this topic and how one operator was addressing it in an earlier blog. Regardless, it is a problem or opportunity depending on how you look at it
Open Source Virtualization
One vendor brought up the issue of hypervisor uptake, particularly KVM as a source for slow rolling virtualization, however, the industry will have to account for operators deploying a mix of VMware, KVM and Xen as well as using public clouds like AWS for part of the service. Advances in virtualization technology have made NFV especially for real-time multimedia Virtualized Network Functions more of a reality, so support for different hypervisor environments will continue to be important.
Start the Journey Now
Are these challenges enough to slow roll virtualization? In fact it’s the opposite. A representative from one of the companies moving towards SDN-based offerings indicated it was important to deploy virtualized infrastructure and get comfortable with NFV/SDN now. Another operator was already underway with proof of concepts leading to commercial deployments for EPC components such as the authentication gateway. Small MNOs were seen to be more aggressive in moving forward with virtualization of infrastructure and services since they don’t have the massive inertia and baggage from legacy services and infrastructure from which they would need to migrate. The cost for moving off legacy TDM for larger operators can be substantial, but as one participant pointed out, it may cost more not to take action when taking into account the extra maintenance, inefficiencies and operational costs to support end-of-life infrastructure compared to virtualized environments.
The Market Reacts
Regardless, one Systems Integrator that has been helping operators formulate RFPs for moving to the cloud observed several trends that they are seeing with operators that they deal with:
- A hybrid approach using public and private clouds
- Seeking independent advice when it comes to cloud orchestration and management
- Use of open source approaches, ISVs, and major NEMs for orchestration
- Development of in-house orchestration and management
What is the business case? Is it a no brainer? One participant noted that when it comes to a green field approach to virtualizing infrastructure, there is definitely a business case. When doing the analysis, they advised the group to definitely take into account the cost of legacy infrastructure support since that cost will continue to rise as more functions go end of life. (Click here to get access to a white paper that quantifies the benefits of NFV and virtualization by PA Consulting.)
Interoperability is Paramount
Regardless of the quantitative benefits of virtualization and NFV, one of the bigger barriers seen by the group was multivendor interoperability. Interoperability is not only an issue between virtualized applications, but also between applications and the virtualized environment (MANO, VIM, hypervisor environments) in which they’re deployed. Even in the presence of standards and specifications emerging from ETSI, OpenStack, OASIS, Open Networking Foundation, Open Source MANO, and OPNFV, interoperability will still be an issue. While issues with interoperability could possibly slow roll the spread of virtualization in and outside data centers, this may be an opportunity for next generation telecommunications providers. Helping operators overcome interoperability issues between networks has long been the value-add of wholesalers and IPX operators. Who knows, in the future, wholesalers/IPX providers may be an enabler of hosted service options and interoperability between operator “clouds!”
Let us know what you think about virtualization in the wholesale/IPX space, and what new service features as well as new services are becoming possible as operators adopt this technology in their networks. You can tweet us at @Dialogic.