There has, for years, been a potential for the cloud providers’ networks to create competition for enterprise networks based on MPLS VPNs. I noted in an earlier blog that enterprises were seriously looking at reducing their WAN costs by using SD-WAN and/or SASE. This obviously generated an opportunity for cloud providers to offer WAN services, and also for someone to offer multi-cloud interconnect. Maybe it also offers more, as Fierce Network suggests.
The increased role of public cloud services as a front-end application element, and the growing tendency for enterprises to think of workers, customers, and partners as roles within a common hybrid cloud application set, has meant that many enterprise traffic flows have already abandoned the MPLS VPNs in branch locations, flowing instead from Internet to cloud to data center.
Google’s Cloud WAN, Amazon’s AWS Cloud WAN, and Azure Virtual WAN are all cloud provider services that aim to increase the role of the cloud provider in what was once a branch-and-HQ MPLS VPN lake. I’m not seeing, or hearing about, any major initiatives from cloud providers to take on branch connectivity missions for non-cloud customers. Enterprises also, so far, seem not to be looking for that sort of relationship, but changes could be in the wind.
Recall my comments on the possibility that something IMS-like might create an opportunity to steer traffic by QoS? Well, these cloud WAN services are all essentially fast paths from the Internet (for access to the sites) to the cloud WAN. If we presumed that there was a fast path from the user to the cloud WAN on-ramp closest to them, it might then create a QoS facility for the cloud WAN that was better end-to-end than an externally provided SD-WAN solution.
Another possibility related to that is the chance that some form of access network partitioning, not unlike wireless network slicing, could offer a way to separate premium services. Would this violate net neutrality? Who knows? There have been times, particularly early on, when neutrality policy also excluded services that aimed at Internet bypass in order to offer paid prioritization. There have been times when it allowed it, and probably times when sunrise and sunset might have been under suspicion. Let’s face it, neutrality policy is in the eye of the party in power.
Another Fierce Network article raises a third possibility. Lumen Technology, who in the first of my Fierce Network pieces was cited in its partnership with Google’s Cloud WAN, is of the view that the Internet is not the enterprise, or the enterprise network. Lumen proposes to build something that is, a fabric if high-capacity Ethernet trunks that link the cloud providers and enterprises. He’d like to connect the local hubs of all these providers too, and I’d not be surprised to find Lumen connecting some major enterprise sites and partnering with metro/regional fiber access players, too.
This is one of those good-news-bad-news things. Enterprises would surely be happy to see cloud providers offer competition in premium business services. Cloud providers would, of course, be even happier to sell them. Telcos? Well, that’s where bad-ness comes in.
Telcos are, as ISPs, more likely subject to neutrality rules, including wavering policies on Internet bypass. In the US, the original justification for extending net neutrality in that direction was to prevent non-neutral services, because they were more profitable, undermining Internet investment that directly impacts…wait for it…voters. The risk, they see, is that cloud providers with no neutrality issues to worry about might end up providing those premium services and increasing the disintermediation the telcos already suffer at the hands of OTTs. Before, they were disintermediated from higher-level, more valuable services. Now they could be disintermediated from services in their own zone.
On the other hand, might operators who used 5G or 6G IMS and network convergence (the much-hoped-for “all-IP network”) offer what I previously characterized, based on some insightful comments, as nIMS across wireline and wireless? Remember that IMS spec calls for specialized bearer channels based on service requirements, and it seems unlikely to me that 6G would deliver on microsecond latency for everything, which suggests that low-latency might be an example of a service requirement for a specialized bearer channel.
This last point may be critical, because it illustrates a paradox that’s been created for the network operators. On the one hand, they’re buoyed by the hope that new applications in general, and new real-time applications in particular, are a major future source of revenue for them. The same thing, I must point out, is the hope of edge computing and many who envision a “smart” future. On the other hand, these applications will surely demand latency and availability guarantees that are not justified for the Internet overall, and may not be economically feasible if they have to be presented on a no-extra-charge basis to everything.
Best-efforts Internet isn’t an ideal, optimal, or perhaps even compliant service on which to base real-time applications. We need to accept, at the policy level, what we’re mandating economically and technically. Real-time is the next time we need to be worrying about. It’s likely that a real-time application would demand a subset of Internet features, optimized for low latency and high availability. Getting to that could happen, in theory, either by refining how real-time-enabling features work, or separating real-time from the Internet. Cloud providers, who are also the most likely players to take a shot at the real-time-dependent edge computing space, have a big incentive to try to advance those special features. Could they do it themselves? Yes, deeper in the network, but not likely in the access network unless the features become part of “standards”.
Could it be? In theory, mobile services (IMS, 4G, and 5G) support multiple data network gateways, so theoretically a cloud provider could offer connection to one of their hubs. Would an operator themselves be allowed to offer something like this? I’m skeptical. This would, in my view, create a reversal of the past policies aimed at keeping operators from exercising an unfair advantage due to their position. Could neutrality put them at unfair disadvantage here?
There’s a potential collision between public policy in the traditional sense, and the health of the market that, to the FCC and many other regulatory bodies, is the primary goal of regulation in the first place. We have standards and technology trying to advance, but at risk to running into a barrier created not by technology limits or economics, but by a policy. The role of the Internet has evolved to be truly the modern equivalent of dialtone, and we need to think about that in technology planning and regulation.
