Cloud computing is heavily promoted as the low-cost alternative to in-house data centers. Yet this is not true for any context and any application. A recent paper with the title “To Move or Not to Move: The Economics of Cloud Computing” by Byung Chul Tak, Bhuvan Urgaonkar, and Anand Sivasubramainam states that especially applications with larger workloads are more expensiveto run on the cloud than in a well-managed in-house data center. The researchers from the Pennsylvania State University ran a battery of simulations, applying the TPC-W and TPC-E benchmarks, which emulate e.g. the workload in an online book store or and online brokerage firm.
Doing their simulations over a 10 year time period, the researchers concluded that e.g. small workloads (20 transactions per second) in their example would cost USD 10.000 to run in-house and USD 1.000 to run in the cloud per year. For bigger work loads this cost advantage can turn into the opposite – an in-house implementation costing USD 400.000 a year can easily surpass the USD 1 million a year level in the cloud.
“Overall, we find that in-house provisioning is cost-effective for medium to large workloads, whereas cloud-based options suit
small workloads,” the researchers state in the paper. This is yet another hint that the cloud is not necessarily a low-cost alternative for every need and should make decision makers aware to carefully analyze their own situation before making high impact decisions.