Now that the cloud is becoming a standard feature in the enterprise, a little truism has emerged: Resources are infinitely scalable, but so are the costs.https://o1.qnsr.com/log/p.gif?;n=203;c=204663295;s=11915;x=7936;f=201904081034270;u=j;z=TIMESTAMP;a=20410779;e=iTheoretically, at least, increased cloud consumption should only happen in the presence of increased business activity, and therefore increased revenue. So the cost/benefit ratio should always favor the enterprise, at least if you’re smart about it. In practice, though, it doesn’t always work that way. But even if it did, the real question is not at what point does a gargantuan cloud presence become a money loser, but when does it end up costing more than building and operating your own data center?
This conflict is particularly acute in rapidly growing enterprises. Companies that go from little-known start-up to must-have business solution provider overnight can suddenly find themselves on the hook for millions per year. Wired.com, for example, tells the tale of MemSQL, a West Coast database services company that originally provisioned its entire test and development infrastructure on Amazon only to dump it one day in favor of in-house, bare metal infrastructure. A simple cost comparison was the key driver: For about $120,000 amortized over three years, the company was able to shed more than $300,000 in cloud costs per year – a reduction of more than 80 percent.
Indeed, the public cloud’s limitations are increasingly showing up in surveys of leading enterprises. Rackspace, for one, found recently that 60 percent of IT leaders view the hybrid cloud as the real end game in their deployment strategies, which may cause the company to rethink its strategy of scaling up its public resources to Amazon levels. It’s worth noting that, after spiraling costs, the next big complaint from users with large cloud presences is the diminishing performance of multitenant resources. Steadily increasing costs with steadily diminishing service levels do not make for a promising future.
Indeed, according Chris Ford, founder of the Chicago Business Intelligence Group, the question many organizations should confront is, “Do I just want be on the cloud, or do I want to be the cloud?” Just as Henry Ford discovered in the 1920s, once you reach a certain size, in-house utility production (in his case, electricity) is always cheaper than third-party sources, plus you can supplement your revenue by selling the excess to willing buyers. Translated to the modern enterprise: An internal cloud should be able to satisfy your needs and finally transform IT from a cost to a profit center.
Some of you may be wondering why the economics of cloud scalability are so off. If the cloud provides 5 TB of storage more cheaply than a typical SAN, then why not 500 TB? Part of the answer lies in the packaging. While it is true that costs for basic commodity services are low in the cloud, those for advanced services like load balancing, app management and file sync/sharing are not. These premium services not only add to the total service cost but often require manual programming and other labor-intensive TLC, and they invariably become more critical to smooth data operations as the cloud environment scales up.
Some providers are fighting this trend, however. IaaS provider Tier 3, for example, recently launched new self-service options to its stack, providing native configuration and management tools, as part of its basic service agreement, that help organizations mirror their on-premise infrastructure. The package includes application load management, virtual LAN provisioning and IP port customization, which company CTO Jared Way says will bring down the entire cloud TCO, not just the cost of resources.
Like all myths, the one about infinite cloud scalability has a kernel of truth to it. The potential physical limitations of the cloud are so vast that the edges can barely be comprehended. But the cloud is more than just capacity – it’s about operations, too. And the fact is that overprovisioning in the cloud can lead to the same cost, complexity and performance issues that plague traditional infrastructure.
Real clouds can theoretically extend around the entire world as well. But if they acquire too much water, it rains.