> Isn’t that a step backwards? It could be for some use cases, but there are significant efficiency advantages to processing the data where it lies instead of

> Isn’t that a step backwards? It could be for some use cases, but there are significant efficiency advantages to processing the data where it lies instead of shipping a copy to separate compute services.