I’ll introduce the decision criteria and reference examples for on-premise vs cloud that I learned about.
Background: Want to Know On-Premise vs Cloud Decision Criteria
I wanted to understand the decision criteria for on-premise vs cloud, so I learned various things from SRE experts.
The advantages and disadvantages of on-premise vs cloud ultimately come down to cost considerations.
On-premise costs:
- Physical server procurement costs
- Data center maintenance costs
- Operational costs
- Performance improvement costs
It takes a large scale before “on-premise > cloud” becomes cost-effective, otherwise it’s wasteful.
As a guideline, the service revenue scale needs to be around 200 billion yen, but this varies depending on service characteristics, so it can’t be said definitively.
For example, Dropbox is a storage service with $1.1 billion in revenue in 2017, but because AWS storage fees were expensive, they stopped using AWS (cloud) in 2014-2015 and built their own data center.
Dropbox Stopped Using AWS / DropboxはAWSをやめた
14年にAWS利用を停止することを決定し、実際に15年にデータセンターへの移行を完了させた。現在はEB（エクサバイト）級のデータをこの自社インフラでさばいている。They decided to stop using AWS in 2014 and actually completed the migration to their data center in 2015. They now handle exabyte-scale data with this in-house infrastructure.
(略) / (omitted)
Cost / コスト
実際に同社の場合、オンプレ回帰によって2年間で7500万ドル（約81億円）のコストを削減できたと報じられている。In Dropbox’s case, it was reported that they were able to reduce costs by $75 million (about 8.1 billion yen) over two years by returning to on-premise.
Cloud → On-Premise Migration
On-Premise → Cloud Migration Examples
That’s all from wanting to understand on-premise vs cloud decision criteria, from a fundamentally cloud-oriented Gemba.