AT&T has built further upon its Network On Demand platform with the launch of Network Functions On Demand, a managed service that enables enterprise users to run multiple virtual network functions (VNFs) from a single server housed on its premises.
On the face of it, the service is NFV in action -- an enterprise no longer needs a dedicated device for each application (router, firewall, WAN accelerator etc) but can add software-based functions that run on a common server as needed. (See AT&T Launches Network Functions on Demand.)
And there's no doubt that the launch of the service marks a breakthrough in service provider capability: AT&T Inc. (NYSE: T) is launching the same service in 76 countries around the world simultaneously (see this press release for the full list of countries) and has four virtual network functions (VNFs) available from day one, with more to follow. AT&T's international experience means it has been able to set up and manage and meet all the regulatory, security and support demands of many local markets.
Yet the service, while underpinned by a next-generation network and support architecture, is characterized by a legacy business model approach that runs counter to cloud models.
The service is enabled by AT&T's Network on Demand network, a distributed cloud architecture built up during the past few years and run from the operator's 74 AT&T Integrated Cloud (AIC) data centers (there will be 105 by the end of the year).
AT&T's ECOMP (Enhanced Control, Orchestration, Management and Policy) provides the management, orchestration and policy management capabilities, while the customer premises equipment is a specially commissioned x86 server "designed to our specifications" that supports KVM and Openstack, explains Roman Pacewicz, senior vice president, offer management & service integration, AT&T Business Solutions. (The operator is not sharing the identity of the server vendor currently.)
Initially, there are four VNFs that can be added to and run on that server: Juniper Networks virtual routing; Cisco virtual router; Fortinet virtual security; and Riverbed virtual WAN optimization. Further VNFs will become available soon and will be determined by customer demand, says Pacewicz, though no details are available just yet.
But while the new service is enabled by a cloud architecture and AT&T is keen to promote how flexible and user-friendly the new service is, it doesn't come with the customer freedom that one might associate with a cloud service, at least for now.
The service is a managed service and doesn't come with a customer portal that enables self-service and end user control: Users need to interact with an AT&T account representative to get the CPE installed, order new VNFs and enact changes.
And the business model seems more old school than new age, with initial standard contracts for Network Functions On Demand set at three years. Pacewicz says there will be a certain amount of flexibility "depending on what a customer wants" but that pricing would then reflect any changes to the standard terms.
Pacewicz, naturally, is keen to point out the benefits that Network Functions On Demand can provide customers, which, he says, have been testing the service and providing feedback for quite some time: The VNFs can run and be managed alongside existing traditional services with the same service level agreements (SLAs); Customers can try out the new service in a small number of locations and with as few or as many VNFs as they want to start and then build from there; and the total cost of ownership of running business-critical functions improves as more VNFs are added, due to the shared resources.
And AT&T plans to share with the industry the specifications of the x86 server it is supplying as part of the service and which has been designed and optimized for an NFV environment: "We see this as an industry development, not an AT&T-specific development -- we don't need a customized approach," states Pacewicz.
But it seems clear that, for a service that sports the term "on demand" and that runs on top of a cloud network, it has a number of restrictions and elements that potential customers might question, including the need for an AT&T-supplied x86 server, a lack of a self-service portal and an initial contract length that ties them in for what seems like a very long time in today's flexible cloud services environment.
— Ray Le Maistre, , Editor-in-Chief, Light Reading