Lots of folks focus on compute and disk cost when they're in the cloud because they're usually the two biggest items on the bill. That's reasonable, but when tr

Lots of folks focus on compute and disk cost when they're in the cloud because they're usually the two biggest items on the bill. That's reasonable, but when transitioning a distributed system to a hybrid infrastructure model (eg: Cloud and a Dedicated Colo) it's important to factor in network cost. Cloud providers usually charge for ingress or egress and have marginal cost for inter-DC traffic (eg: availability zone to availability zone on AWS). Distributed systems are chatty by nature, so if requests are constantly leaving your cloud and entering your new DC you're potentially paying twice for the same transaction. This cost adds up fairly quickly. The same thing will occur if you operate an application in two regions on AWS and have a lot of region<>region activity.