we are using k8s that automatically generates Load ballancers when using ingress. In our use case, we are using TCP/WebSocket connection with each client app, so that current Load Balancer default limit that is set to 10 000 concurrent connections is easy to reach.
I know, that DO supports 3 levels of Load balancers, the higher available allows up to 40 000 connections.
From our point of view - we don't want to micro-manage load ballancer status and manually upgrade / downgrate it. It is even not possible (now) to downgrade / upgrade load balancer. What we want is automatic scalable load balancer, that upgrades / downgredes itself as needed. In most days, the basic load balancer is good enough for us. But peak can happen and we cannot currently react on it. The load ballancer is bottleneck in our architecture. We can easily add new nodes to k8s in minutes, but we cannot easily to the same with load balancer.
So 2 requests :-)
make load balancer reconfigurable -allow us to easily migrate between LB plans without IP address change
give us some starts - we want to know, how much connections is used to predict upgrade / downgrade
make "autoscaling" load balancer. Easilly - for example when 90% of connections limit is reached, upgrade. When less than 40% is used for few days, downgrade.
give us ability to use load ballancer that handles 100 000 concurrent connections or more.