Let me explain. A lot of people talk about multi-cloud these days. Either AWS and Azure, or AWS and their own datacenter, or whatever.
While it's really easy to send compute jobs to one cloud or the other based on which one is best suited for the job, the big issue is that your compute job will need data. Right now your choices are to have a copy of all your data in all the clouds (a very expensive proposition since you'd be constantly shipping updates across very expensive outbound connections) or to have all your data in one place and have the compute jobs reach out across the internet to get it (also expensive and now there is latency for every job too).
I want a proxy that is smart enough to say "Workload X is usually done on AWS, and requires data points Y and Z, so make sure Y and Z are always up to date on AWS, but lazy update Y and Z to GCE in batches through the cheapest direct connect possible".