The Changing Landscape of Global Network Services

The landscape for World-wide network connectivity has been dominated by traditional carriers. Names like AT&T, Verizon, and BT. Although some of of the names may be relatively new (e.g., TATA, Orange) based on mergers, acquisitions, and business model expansions (e.g., moving from a regional to global model), the approach to providing services has generally followed a similar pattern. These are the traditional “telecoms”.

Driven initially by voice communications, the World was slowly, and then more rapidly connected by a series of every more capable Submarine Cable systems. The next phase was driven by multinational companies’ (including so called Enterprise customers) demand for data transport to meet their corporate needs and connect to their supply chain. With more demand, carriers responded by expanding their Points of Presence (PoPs) to new cities and creating consortia to build new Cable systems to increase network capacity and diversity.

The penultimate stage (at last from the perspective of where we are today), is the incredible impact of the Internet. With corporate data centers holding the family jewels located at private and commercial colocation facilities around the World, Internet traffic increased at unprecedented rates of growth driving facilities investment for colocation and hosting space as well as new consortium cable systems (as well as system upgrades based on advancements such as coherent optical technology). These networks, were (and are) provided by the appropriately named “Internet Service Providers” or ISPs.

The underlying business of the typical ISPs can been roughly seen by putting them into two groups. The first are relative “pure play” ISPs. These are companies that derive the majority of their revenue based on terrestrial communications, and therefore global expansion is part of their revenue expansion plans. The second group has more complicated revenue models, where the vast percentage of their revenue is based on their mobile wireless services in their home country (or region of countries). The reason this is important is the emphasis that these companies place on different aspects of their business.

The first group must expand their network services to attract commercial customers to their network. They have to show value and provide the customer expected high-touch support. The second group is at a crossroads. A large multi-country network may provide several millions of dollars a year in revenue, however it comes with associated high-cost to provide service. On the other hand, generating another 50,000 wireless customers (or wireless Internet of Things - IoT devices) will have the same effect, without the high-touch customer required, and it provides additional revenue for wireless services evolution (e.g., 5G upgrades).

While the scenario between ISPs plays out, a new set of companies with unprecedented growth is changing the landscape - these are the Cloud Service Providers (CSPs) - Amazon, Google, Microsoft, etc. It is now time for large global network consumers to evaluate new options for enterprise network services.

The typical customer understanding and use of CSPs is for their processing, storage, value added services, and Internet access. Thousands of companies have either moved part or all of their processing and storage (a.k.a., hosting) needs to CSPs. The major CSPs are seeing 40% revenue growth per quarter and revenue now runs billions of dollars a month. This vastly outstrips the growth of traditional corporate enterprise network growth and total revenue of the ISPs. This reality has created two effects:

• CSPs have created their own network infrastructure, building nationwide networks, consuming huge amounts of existing, and now building their own, Submarine Cable systems (more on this later) - all of this to handle something on the order of 50-70% of all Internet traffic with extraordinary network diversity and robustness
• World-wide and regional ISPs have responded to their Enterprise customer’s need to use CSP resources by extending their ISP networks, bringing Internet peering and MPLS/VPN services directly into CSP facilities

The result of this is that CSPs are becoming the most capable global network infrastructures by far. As they expand in scale (more capabilities at existing CSP locations) and scope (more CSP locations), this capability will only increase. Feeding this monster, that is getting customer access into place, will be an ever expanding set of capabilities from local, regional, and national wireline and wireless networks.

So, what does this mean to the large global company that needs robust and cost effective communications around the World? It means that there are new technical and business approach options that need to be considered.

On the technical side, these companies need to:

• Map their geographic network requirements against the major CSP data centers (regions)
• Understand the CSPs inter-region services and their cost structure
• Understand the global and regional network providers that have a PoP in the CSP regions
• Develop and approach to leverage the CSP’s virtual services to develop an Enterprise network backbone that can use the network services at the CSP’s location (including Internet access)
• Develop an approach to securely leverage multiple local access providers by using Software Defined Wide Area Networking - SD-WAN. This includes regional MPLS/VPN providers (e.g., MPLS/VPN), 4G and emerging 5G wireless, Internet services, and satellite services and integrate into the developed Enterprise network backbone
• Understand each Enterprise site’s service needs and cost trades:

1. How much bandwidth?
2. What service resiliency is required?
3. How long a service interruption can be tolerated?
4. Make the trade between expensive MPLS/VPN and cheaper Internet bandwidth (see Yikes, Internet for Enterprise Services)

• Map out the regional networks that could support each site’s service needs

On the business side:

• Understand the opportunity of leveraging the CSP for the Enterprise network backbone combined with the traditional Cloud services to meet the Enterprise’s needs
• Understand the complexity of performing as your own multi-vendor integrator, buying network services from multiple carriers to provide Internet and MPLS/VPN services needed to connect to the CSP-based Enterprise backbone

In summary, this is a time of significant change all every level of the Information Technology “Stack”. From how to operate and build a network (using Software Defined Network - SDN orchestration and SD-WAN), virtualization of devices (using Network Function Virtualization - NFV), and network resources, to how to balance Enterprise on-premises hosting with CSP-based services.

Understanding these new capabilities and navigating the complexity to create the high-performance and cost-effective network services for an Enterprise to be globally competitive is the challenge.

Buckle-up, all infrastructure is software and your enterprise needs enterprise orchestration

The capabilities landscape for equipment manufacturers, service providers, and enterprises is rapidly changing, and within several years it will be fundamentally transformed from today.

There are several basic areas:

1. The creation of robust commercial Cloud services with a rich set of services all presented for allocation and configuration to the enterprise via a set of standard Application Programming Interfaces (APIs).
2. The emergence of the Software Defined Networking (SDN), offering the potential of flexible network services again presented to the enterprise for allocation and configuration as a set of APIs.
3. The transition of traditionally physical network-related devices to application that can be configured onto essentially standard computer servers, called Network Function Virtualization (NFV).

The battle that is the force driving these changes are between what I call the “new traditional” service providers and the “legacy” service providers.  Companies like Amazon and Google eschewed traditional wisdom of hardware providers and the paradigm of legacy service providers.  Driven by their application development and low-cost consumer mindset, their general approach is to strip-down to the necessary hardware and software functionality.  Bloated hardware and software with features and functionality not needed is removed.  The over 30 years of the evolution of Internet standards that defines to the control of network devices, embedded into expensive routers and switches, is discarded in part or whole for so called “white box” hardware and Open Source software as the basis for their control.

Legacy network service providers grew-up with the Internet, driving its standards within the common framework of a set of “autonomous systems” configured by the service provider with a set of defined end-user services.  Scant thought was given to providing end-users (in this case the enterprise customer) any meaningful end-to-end control of services, and almost without exception nothing that looks like a web-service RESTful API.  This is in stark contrast to the rich information and control APIs expected and provided by today’s commercial Clouds.

The figure below represents the recent past and much of the present.  Blue represents the legacy infrastructure approach.  Focusing on the network space, the enterprise has to contend with complicated device configurations and essentially static service configurations from their network services provider.  There is little if any coordination between the network and the applications development and operations environment other than at best service tickets and at worse verbal (and undocumented) direct staff-to-staff communications.

The expectations of enterprise Information Technology organizations will also drive the trend to a more software defined environment, as the use of Cloud services and it associated reporting and control will become the expectation, not the exception.  In fact, it is likely that more comprehensive “enterprise orchestration” systems will be developed that will cover all services, from internal application development lifecycles (i.e., Development and Operations), to control and management of end-to-end enterprise services delivery.

This leads to the view in the figure, below.  Red and green represent the new infrastructure trends and blue represents the legacy environment.  The significant change is that nearly all of the infrastructure is now software based, from SDN controlling and reporting of end-to-end network (including to and from Cloud resources), to the direct control of virtual network devices whether in the Cloud, at an enterprise location, mobile, or one of those Internet of Things devices using NFV.

When every resource or service is controlled by what appears to be a web-service and the same mechanism is used to obtain performance, usage, and other relevant from across the different traditional service domains (compute, storage, network, security, etc.) then everything looks like software.  Once this happens, one has to completely rethink an enterprise’s IT operation, as the same types of activity that is done to develop applications is now the fundamental discipline for orchestrating the enterprise, whether it is resource management, application development or rollout, or cyber security.

Buckle-up, time to become an enterprise orchestration programmer.

Software Defined Network (SDN) Considerations for Commercial Network Service Providers

More than four years ago I wrote that carrier service providers must realize that they are in reality “application enablers” see (Going Horizontal on the Vertical) and that Service Providers needed to Virtualize the WAN.    These posts indicated two important areas that relevant to the discussion of Software Defined Networks (SDN) in the context of commercial services (A good summary of describing the SDN approach as compared to the prevalent distributed control model can be found at SDN A Backward Step Forward).

The two areas I described were that:

1. Network service providers have to provide an Application Programming Interface (API) for service requests and service status
2. Virtualization of the Core network enables customers to define what they want out of their network ensemble. 

It appears to me that the evolution of SDN technology needs to follow a similar path as virtual machine technologies.  That is the maturation of technology at the enterprise-level and then the transition to a commercial services provider’s infrastructure.  The first instantiation of virtual machine technology enabled the consolidation and increase in efficiency of an organization’s infrastructure.  The next steps were the development of multi-tenancy environments and transition into commercial “Cloud” services.

The successful public service provider SDN controller must be able to provide a user with the control needed by their application set, while at the same time enabling the service provider to optimize their network.  There are significant options about what the SDN concept means for a service provider and what it looks like to a customer:

1. Is the goal to provide “raw” virtualized standard router-type services? That is, does the customer select a router type and instantiate a set of virtual core routers to meet their requirements?
2. Is there a network equivalent of an x86_64 (i.e., Generic PC processor) virtual machine?  Do you provide a blank sheet and development environment for customers to create their own virtual network devices?
3. Is the goal to make it appear that the network is a single large “router” regardless of the number and physical locations of the actual network?  Wow, I get my own world-wide network that looks like a single router!  Can you also provide a “primary” and “backup” virtual router?
4. Do you provide a Plain Old IP VPN service for customers that just want the same-old-same-old basic service?
5. Do you provide multiple network personalities at the same time to a customer?  That is a network connection and control that enables both IP transport that appears as standard WAN VPN services (with traditional standard QoS – 99.9% packet delivery) as well as services that are more tailored data center operations such as moving Virtual Machines (with near Time Domain Multiplexed QoS – 100% packet delivery). 

As the “simplicity” increases from the customer perspective, the complexity for the service provider increases to manage the increasingly dynamic nature of the services offered and customer demands.

Finally, can we expect novel SDN-based capabilities to be provided by traditional network service providers, or do we need companies that “think outside the box” to move into this area?  If the introduction of large-scale Cloud services define the pattern, then companies like Google and Amazon may lead the network charge.