The path to the cloud isn’t always smooth. A recent study commissioned by iland and conducted by Enterprise Management Associates (EMA) showed that when it comes to Infrastructure as a Service (IaaS), enterprises still encounter a number of challenges, some of them unexpected.
IaaS frustrations include scalability, performance, pricing
From the beginning, one of the most-hyped features of cloud has been its scalability, but the reality doesn’t necessarily measure up to cloud providers’ promises. In talking with administrators, EMA discovered that thirty-three percent of them didn’t find scaling up and down to be as easy as expected.
Lilac Schoenbeck, vice president of product management and marketing at iland, believes there are likely a few reasons behind these rather surprising findings. “At its very basic form, we think cloud should be automatically scalable,” she said. A website selling the season’s hottest product is a good example. Traffic can explode very quickly, and the need for near-immediate scalability to address that comes with obstacles.
“First, it’s very difficult, in an automated way, to determine whether something is a spike in demand for your website or your application, or whether something is a malfunction of the system,” Schoenbeck explained. A denial of service attack or similar issue could initially appear to be an increase in demand, and a mechanism that automatically scales in response may not be a good thing. Fast scaling could, in fact, end up being detrimental.
In addition, quickly adjusting compute or other power may require work on the part of the cloud provider. Just because the enterprise’s internal team isn’t the one making it happen doesn’t mean someone doesn’t need time behind the scenes to carry out the necessary administrative changes. In some IaaS cloud environments, Schoenbeck said that notifying the provider about an anticipated increase in demand doesn’t always result in immediate scaling. Instead, she said the provider might say, “‘In order to grow the footprint of your website or your app, you’re going to have to shut it down and move it to a bigger space.'” That means downtime that many customers don’t expect, and in some cases it can cause a significant business disruption.
Customers’ relationships with their cloud providers can also run into trouble when either the vendor or the customer (or both) inadvertently overload their systems and hinder performance. “It’s very easy to do in a virtualized data center,” warned Mike Meyer, chief information officer at Honolulu Community College – University of Hawaii. In EMA’s study, thirty-eight percent of respondents said cloud performance had been an issue.
Meyer, who also has experience in cloud provisioning and has helped a number of organizations transition to cloud environments, explained, “You can get overwhelmed with virtual machines if you aren’t careful, because it’s so easy to set one up but forget to take it away.” This sometimes happens on the provider side, when a vendor doesn’t keep pace with infrastructure upgrades on the back end as they add customers, or on the customer end if one organization has an unnoticed proliferation of virtual machines. The entire customer base may see decreased performance as a result.
Another sticking point with study participants—though this one is perhaps less surprising—was price. Thirty-eight percent said IaaS costs could be confusing or unpredictable. Bruce Templeton, CEO of NephoScale, Inc., agreed some providers have price structures that are “too complex for people to understand the billing, and they get nickeled and dimed and it’s hard to budget around.”
No matter how sophisticated customers are, Templeton said it’s likely they would appreciate price tiers that are easier to grasp and that provide for more predictable costs. He anticipates that cloud providers may “get as creative as they can to make it easier to understand and digest and work with” to counter this concern. IaaS price structures currently follow a wide range of models, including pay-as-you-go, monthly, membership and reserved.
Disconnect between customer expectations and cloud provider offerings
IaaS providers are, in some instances, still trying to adjust their practices to meet customer demands. Templeton pointed to the segment of vendors who have not historically been in the business of supporting enterprise customers, using Google as an example. “They don’t usually give you a phone number to call,” he explained. “They’re designed to do everything through the Internet and ecommerce platforms.”
Administrators, on the other hand, want a different support experience. Eighty percent of administrators in the EMA study said they require some amount of professional services, with forty percent listing ongoing management as one of those services. Those seeking help with security or compliance services, disaster recovery planning and integration with a local data center ranked even higher.
In many cases, both sides—enterprises and IaaS providers—could be doing things differently for a better experience. Customers’ expectations often start out a bit high, but those beliefs may be encouraged by providers selling the cloud services. “You then immediately have very tall expectations for something that is still relatively new,” Meyer said.
Exacerbating the issue is the usually limited amount of knowledge many administrators hold about the technical side of delivering cloud services. “People don’t have an adequate understanding of some of the broader things involved, to know where it can be difficult and where it’s easy,” Meyer said, adding that it echoes what many administrators will recognize as an old IT problem: People think some things are very easy when in fact they’re quite difficult, and vice versa. Even experienced enterprise administrators often lack a detailed awareness of the technology challenges behind cloud services.
IaaS: What enterprises administrators should know and what they should ask
Having a successful cloud experience often requires that enterprises dig a little deeper into how the services will work and what they should realistically expect. Meyer said it all begins with making the right inquiries. “If a user is going to buy services, they should be asking questions like how many base servers are you running, how many are in a cluster and how many clusters do you have?”
This may or may not produce useful information, as some cloud providers hold that sort of data very close to the vest. If administrators aren’t able to get a satisfactory answer from a particular vendor, Meyer said, “I would look for someone that will give you that kind of visibility into their infrastructure, because their infrastructure produces the environment you’re purchasing.”
Before launching any sort of cloud initiative, Templeton encourages enterprises to carefully and honestly evaluate the amount of support they need and are willing to pay for. “We’ve noticed that anyone spending over about $1,000 a month with us wants to talk to us,” he said. Unlike some early expectations, where many vendors assumed customers would simply plug in a credit card number and start scaling their cloud services without ever asking a question, Templeton said of his organization’s customers, “They want to talk. And they should, because IT is that complicated.”
Photo courtesy of Shutterstock.