Like any tour de force that is fast, flexible and dynamic the cloud can create all kinds of problems if you don’t tame the beast. It is a phenomenal weapon, but without the right training, you can point it in the wrong direction and cause carnage
There are many potential problems, but they all fall into one of five most commonly recognised categories: expectation, litigation, integration, inflation and orchestration.
Expectations
Many who start their journey to cloud computing have the wrong expectations about what cloud systems can do, about what kind of applications should run in a cloud, about the skills required to operate a cloud, about the cost of a cloud and the length of the journey to maturity, according to Alessandro Perilli, general manager of Red Hat’s open hybrid cloud.
The expectations of building private and hybrid clouds are frequently wrong, says Perilli, as is the perception of consuming public clouds
But it’s an easy mistake to make, even for a CIO, says Perilli. “Most of them are not obvious. You need to acquire expertise and experience to properly assess the challenge and define the best cloud adoption strategy.”
Litigation If you are starting from the wrong point, then any service level agreements are going to be built on unstable ground too. If your expectations are wrong, how on earth can you tell your lawyer to draw up a contract to cover eventualities you are oblivious to?
Ideally, the cloud services provider will make it its duty to make sure its clients know enough about cloud services, so they understand the limits of their contracts, says Simon Walker, managed services director at cloud service provider Softcat. But how often does that happen?
As so often happens projects, assumption is the mother of all cock-ups and it would be easy for a service provider to assume a CIO knows all the legal ramifications. “We've had a few instances where customers have misunderstood aspects of the contract from other suppliers and what they're signing up for,” says Walker.
This could be avoided in the first place through better education, he says. The best way to get knowledge is to keep asking questions, even the ones that seem stupid. There are no stupid questions, only stupid corporate legal fees, so get our embarrassment in early. Unexpected costs are the next most common problem.
Inflation All kinds of factors contribute to galloping costs. Many IT heads readily admit they don’t know the scale and type of cloud use, according to Phil Turner, EMEA VP at Okta.
That, in turn, will lead to software licensing and governance problems, as the pay-per-use culture of the cloud model isn’t always as straightforward as it appears. How many users ever read a software licensing agreement before purchasing? Have you ever got to the end of one?
You only need to look at the differences between enterprise and service provider licence agreements, or how Oracle works in a multi-tenanted environment, or Microsoft’s complex licensing and complete lack of strategy to make the system work in a cloud world to see how confusing it is for customers,” says Stefan Haase, product director at cloud provider Redcentric. It even confuses the service providers, he says.
Another cost to be factored in is over the endless dilemma over proprietary and open systems. Yes, proprietary technology is expensive, but few people realise that open platforms – the supposed liberators of the locked in CIOs - can cause unexpected budget losses too. “There’s massive enthusiasm for OpenStack now but the problem is there is a huge development effort required and tons of security decisions to make,” says Randal Asay, CTO at cloud security vendor Catbird, “And if you know anything about development communities, they are not known for their security policies.”
One of the most common problems with the cloud is the lack of visibility it presents,” says Okta’s Turner. That nebulousness also contributes to next most common issue.
Orchestration Cloud computing often lacks the liquidity to give you a perfect internal market for your computing resources. That’s because there isn’t perfect communication and completely free movement of
the means of production (the computing capacity, memory, storage and networking.
Clouds are designed for scale and elasticity and typically use cheaper commodity components. That’s a potential problem, says Dave LeClair, senior director of strategy at Stratus Technologies.
To keep cloud applications up and running, they need to be designed to work around potential failures. For existing applications that have been tried and tested over years, this means a total rewrite to make them cloud-ready.
This is not only a costly and time-intensive exercise that requires specific skill-sets, there is also a significant amount of risk attached,” says LeClair. Applying those time-honoured approaches to availability in the cloud doesn’t make sense. What is needed is a cost effective technique that uses the inherent elasticity of the cloud to provide the availability expected by individual applications, at specific times and under specific circumstances.
Software-defined availability is the next generation approach that satisfies this need for availability, in this always-on world, underpinned by this next generation of computing,” says LeClair.
Migration CIOs are often told to cut costs and move workloads into the cloud, says Ash Patel, Insight UK’s head of cloud for EMEA. This leads to another problem.
As you add more cloud services the CIO should consider the management questions that will determine how cohesive the IT infrastructure will be. How do you integrate multiple clouds into your current environment? Do you add single sign on? Can you? How do you get one cloud provider to connect to another? Is there one out there that can do it all? “Managing these services becomes very complex, both from an infrastructure perspective and user,” says Patel.
The marketing talk says cloud makes it easier to manage your service. This is true for a cloud service in isolation, sometimes, but most businesses don’t stop at one service,” says Patel. So the big picture gets complicated – and complicated is expensive. To configure the whole IT structure, both cloud and on premise, so that everything is working in harmony, sounds a lot easier than it is.
So how do you choose the right cloud service? Is it better to re-engineer your current on-premise environment to the cloud? Or do you work with what you have and try and choose the cloud services that are easy to integrate? How will you manage your hybrid environment, how will your IT work with cloud services?
These are all questions people ignore in the rush to get into the cloud. But the biggest mistake you can make is ignore the details as if they are some trivial software licence agreement.
In the cloud “the contract negotiations can take longer than the estimated project duration,” says Lori Williams, Appirio’s general manager for EMEA. But is that really a bad thing?
To avoid the most common mistakes:
- Do have realistic expectations of what cloud is going to bring.
- Do read the small print: make sure you fully understand the ins and outs of the service before you sign
- Budget carefully: on-premise and cloud are different beasts and you need to look closely at details on licensing
- Be prepared to rein in services that aren't needed; monitor expenditure closely
- Make sure that you understand the implications of downtime and prepare for failure
- Don't stint on lawyers' fees or worry about negotiating SLAs – getting it right first time will save a lot of grief later
Edited on 30 July to correct some wrongly attributed quotes
