Like a bad Internet meme, I often see assertions that public cloud storage costs are higher than in-house storage over time, making cloud unsuitable for long-term data preservation. A recent article on StorageSwiss made this argument too, concluding public cloud is not ideal for long-term data storage.
Conspicuously missing each time I've seen such claims are any supporting numbers. Why?
We at HubStor wholeheartedly believe the opposite. And not because of an emotional reaction or bias. Simply run the numbers!
Public cloud economics - even for long-term storage - are unmatched by most in-house storage implementations. Arguments to the contrary are most likely giving you an apples to oranges comparison.
Apples to apples
Comparing traditional in-house storage with the cloud isn't simple - it's apples and pomegranates. Because the two models are drastically different, we recently published a paper comparing HubStor's all-in costs with the fully-burdened costs of in-house storage. (requires registration)
Notice I said 'all-in' and 'fully-burdened'. We're looking at the Total Cost of Ownership (TCO). We often see the unfair comparison: cloud with all its variables compared only to the upfront acquisition cost of in-house storage, leaving out the full-burdened picture of storage costs in-house. It's a much different picture when you factor a hardware refresh every 3-5 with data migration costs, and all the overhead storage needed for data protection and the desired usable capacity.
What our numbers show
In our TCO model for in-house we calculated costs for 100TB of usable capacity amortized at 4 years, arriving at $398,438 annually (or $0.32 per gigabyte monthly). That's lower than ESG Global's and significantly lower than Forrester's, which we wrote about previously. And it's much lower than the reality I hear about directly from storage admins that pay attention to their fully-burdened storage costs.
Our model, which we attempted to normalize on this same 100TB, showed an active-archive HubStor tenant with high performance compute at $98,628 annually (or $0.08 per gigabyte monthly). A cold archive configuration for the 100TB in HubStor where compute is scaled down comes to $22,000 annually. Even at the high-end performance scenario with an active archive configuration, HubStor is 75% less than in-house storage!
Remember too that we are comparing HubStor's all-in pricing which includes the public cloud costs.
Indeed you get a very different picture when you compare fully-burdened costs instead of just the upfront capital expense for in-house storage. In year 5 or 6 with HubStor you could still be paying <$100k annually for the 100TB; whereas, by this time with in-house storage you'd have incurred another large capital expense for the new hardware and a data migration project for your 100TB.
How much of a factor are cloud's data transfer costs?
Another often cited "gotcha" of cloud storage is data transfer cost (retrieving your data). It's certainly a variable in cloud economics - one that doesn't exist with in-house storage - so in our next post we'll quantify it for you.