What is behind their distrust? We’ve recently seen significant outages in major cloud providers due to configuration and policy errors from Tier 1, 2, and 3 service providers. Network reachability failure occurs due to a few reasons.
Figure 2: Tier 1 is being replaced by the cloud provider backbone.
click to enlarge
Larger providers understand these problems very well, especially with small carriers. Local issues sometimes cause global outages. For example, if your local ISP (Tier 3) has one of the aforementioned routing problems, your connection to the large provider (e.g., Facebook or AWS) will fail. Essentially, the ISP promises a destination or service that it can’t deliver. They don’t have end-to-end control or visibility. They have to trust that routing rules propagated from higher tiers are accurate, but as we’ve seen, that is not always true.
We can’t afford the current model’s internet-based connectivity for mission-critical services. On the other hand, if we change a few things, which in my opinion and experience is critical for next-generation business class internet, we can establish direct control plane peering between the end customer and the cloud provider.
We start by introducing a Network-as-a-Service provider. This NaaS provider sits between the enterprise and the provider (Facebook, AWS, etc.). The NaaS provider has no transit information about customers and cloud provider routes. This means