Respective CSP network functions have provided a bridge between virtual private clouds and networks however, these functions are myopic in their delivery – built only for the cloud in which they were conceived.  Even within a single CSP infrastructure, these tools lack the global visibility that most network engineers are accustomed to in order to effectively manage the network.  As organizations continue to scale their cloud infrastructure across regions and other clouds, they are faced with either poor service delivery due to a disjointed network, or the risk of security vulnerabilities created by connecting everything through a wide-open, peering mesh. 

While cloud-native network functions provide the most efficient access to the network below a CSP’s abstraction layer, the requirement to manually stitch together multiple CSP tools is extraordinarily complex and unwieldy to operate, manage, or scale.  While CSP network functions have matured, most organizations still require complementary tools and services such as load balancers and firewalls; this only serves to add additional complexity to an already complex problem as well as increase the risk of technical debt.

Many organizations and vendors have attempted to rationalize cloud networks by employing orchestration solutions that replicate their on-prem network through virtual versions of conventional network appliances.  While this may help in connecting cloud networks and application endpoints, it only serves to create unnecessary complexity.  The “allow all” approach of conventional routing policy only serves to break the existing network segmentation and security of cloud-native networks.  Unfortunately, regardless of the approach, all are inherently complex and do not solve the key issues that make multi-cloud networking extremely painful.