My grandparents worked with mules. While few contemporary readers are familiar with these crosses of male donkeys and female horses, US farming in the early twentieth century included a crucial role for six million (!) of them.
Keith Townsend says so in his recent "Why there's no compelling use case for hybrid cloud", and he's right. The original aim was that a modest private cloud could overflow its bursty loads to the unlimited capacity of the public cloud. The private cloud would form a less-expensive baseline, and peak demand could pay the higher costs of peak capacity.
That sketch excluded too many crucial parts of reality. Creation of a "single pane of glass" to consolidate views of private and public resources has often--always?--proved to be more expensive than the computing it manages. The consequence: organizations end up assigning loads exclusively inside or outside their perimeters. The only operational hybridization they exhibit is between the two managerial regimes.
Failed hybrid operation
Hybrid cloud is a story. Balkanized private and public management is a story. Which story is true? It's not yet certain: IBM, HP, and plenty of consultants continue to advertise hybrid successes, while Townsend, at least one IBM consultant, Lilac Schoenbeck, and other reputable commentators testify the cases they see all turn out to be mirages. Maybe hybrid cloud is so wonderful that the successes with it are keeping it a secret; that can happen.
My hybrid-cloud conclusions as we near the end of 2016:
- Hybrid is not yet standardized best-practice. If someone tries to sell you hybrid, think through the alternatives and risks carefully. Hybrid cloud purchase at this point is closer to microsatellites than paper towels.
- Public clouds are hustling to solve objections based on cost and compliance.
- While computing use is inherently bursty, public-cloud recruitment is far from the only solution. A true private cloud necessarily implements chargeback. Private clouds substantial enough to consider hybridization are simultaneously big enough at least to experiment with peak pricing.
- Load migration is real--at least, on an OpenStack basis. For a public demonstration, read about the most recent InterOp Challenge.
How does the hybrid cloud look from where you stand? Let us know in the comments below.