This article originally appeared in the Jan. 2012 issue of Cloud Computing.
Automobiles make it easy for an ordinary human to propel themselves forward at a relatively high velocity. Although they’re easy to operate, the technology upon which they’re built is incredibly complex. Cloud is no different in this respect.
Technology providers of every stripe are rushing to deliver a cloud solution. Some of these providers have historically limited their business models to the provision of tangibles (i.e., power, bandwidth, and/or rack space). Others have grown a sprawling hardware monoculture heavily reliant upon “lock-in” management tools and vendor-provided engineering support. When attempting to adopt cloud technology and offer cloud products / services, it seems that many expect to cleanly shift these existing business models onto the cloud.
However, the times are a-changin’. Here are a few key examples of “under-the-hood” complexity that turn traditional IT business models on their head:
1. Hardware: In the past, a typical server failure resulted in a fairly substantial repair exercise. A new server would have to be brought up using an appropriate OS image; the basic environment would require some tweaking to be production ready, data would have to be restored, connections to storage facilities manually rebuilt, etc. Ultimately this meant that expensive, high-end server platforms were critical to the success of any large deployment and therefore justifiable. Losing three to four servers a month in a large-scale deployment would endlessly tie up IT administration resources.
In cloud, the components of a compute environment are separated into pools. Losing part of a resource pool simply means that the corresponding load is shifted elsewhere. Since processing, memory, and storage move about freely within a cloud environment, hardware transience becomes significantly less detrimental to operations. Spending tens of thousands of dollars on high-uptime hardware with pricey service contracts just does not make any sense when hardware persistence is no longer an issue.
With smart players adopting commodity hardware strategies for their cloud offerings and nimbly adjusting to the resultant shift in support concerns, I am not sure how organizations dependent upon big-budget hardware platforms will effectively compete.
2. Software: Tight integration and streamlined management will win the game. Home computers were originally shipped as kits. Who, but die-hard hobbyists, bolts together their own PC’s anymore? Originally, the only way to build a cloud was to bolt together various pieces of software (hypervisor, OS, management, monitoring, etc.). The home-brew approach is no longer necessary or advantageous—yet some continue to do it. If you intend to offer cloud services which require wrenching-together a dozen products from disparate vendors (each with a unique take on software licensing), you’d better tighten all the nuts securely because if one spring comes loose, the whole house of cards comes crashing down and you’ve just entered the twilight zone of vendor finger-pointing.
3. Monitoring/Support: What does your NOC (News - Alert) currently monitor? Circuits? Layer 2/3 network devices? Infrastructure, storage, and host memory/CPU stats? Backup status? If your answer to all of these was “yes” then consider yourself prepared for this aspect of cloud delivery. If not, don’t think for a second that clients will sign-up for cloud services with disjoint or partial monitoring. A nice pretty screen at the front of the NOC with network ping statistics just won’t cut it any longer. Since cloud can envelope the client’s whole infrastructure, you will be expected to not only deliver the goods but also monitor every aspect of the stack you’re delivering.
Similarly, if your existing support staff has been trained to deal with power outages and telecom issues, don’t be surprised when they’re overwhelmed by the inevitable and painful “my (virtual) server seems to be running slow” calls. Yes, as a budding cloud provider, you now own all of that too.
4. Billing: Since you’re now a ground-up infrastructure provider, your customer will quite reasonably expect to receive a single bill for services. Did you think you were going to just outsource the monitoring and support pieces? Perhaps you thought the customer would willingly carry separate hardware costs just for the privilege of using your cloud service? Maybe the crazy notion of signing customers up with a 3rd-party for network connectivity has crossed your mind?
Your cloud solution is not going to get a warm reception if you point your customers at peripheral service providers to fill the holes in your solution. Currently, customers equate cloud with comprehensive. Back-end processes are expected to be every bit as streamlined and integrated as the core technology. If you dare breach this expectation, you will watch your potential sales wither on the vine.
This is not to say that things cannot be outsourced or otherwise contracted but your vendors must understand the imperative in working together on a seamless and cohesive solution. Not everyone is going to get their name, fancy corporate logo, and whiz-bang GUI plastered in front of the end-client. Also, be prepared to spend some development dollars customizing a management portal for your customers. With all of this integration, some degree of custom development is inevitable.
Of course there are many other considerations when designing a cloud offering. Experience has shown that these items cause the most surprise and/or consternation amongst newcomers. Everyone benefits when these issues are addressed during the product development cycle and businesses arrive to market with mature, responsive cloud products.
Edited by Stefania Viscusi