* Cites management practices to consider before moving applications to the cloud
Everyone’s talking about putting their application “in the cloud” these days. Cloud computing might be the next evolution for enterprise applications, but BMC Software’s CTO advises IT organizations to consider how their application can be managed once in the cloud.
I recently had the opportunity to talk to Kia Behnia, CTO at BMC Software. Given that BMC’s core business is software to automate, control and manage enterprise systems, I thought it would be worthwhile to get Behnia’s insight on managing applications in the cloud. “Cloud computing” is the buzz phrase of the day, and Behnia reminds us that there are important management practices to consider before and while moving applications to the cloud.
FAQ: Cloud computing, demystified
“Two year ago, BMC made virtualization — and by extension, cloud management — a corporate strategic initiative,” Behnia says. “We talked with several of our large enterprise customers, as well as large service providers and outsourcers, about their plans for adopting virtualization. We talked about what they saw as their challenges in regards to both adoption and management, and where they saw their vision at that time. Many of them talked about self-service, and being able to dynamically provision systems. This was prior to the term ‘cloud computing.'”
Behnia says these discussions led to BMC engaging with these enterprises on a number of their virtualization projects. “One of the critical things we realized was that virtualization, which is really the precursor for ‘the cloud,’ amplifies the need for having a unified integrated management architecture, because the environment is shifting radically and too quickly becoming dynamic. It’s very difficult to keep disparate [management] point products up-to-date with what you have, what’s going on and what the impact is,” Behnia says.
“We invested quite a bit of our resources, not just in adding functionality that was very specific around virtualization management, or cloud management, but also investing in the integration layer that is required for both BMC and non-BMC products to interoperate and integrate and work together to provide a broader set of management capabilities,” he says.
The heterogeneous nature of cloud computing is one of the aspects that draws customers to it. No one wants to get locked into a particular vendor’s solution or a specific computing platform. “Virtualization gives customers choice,” Behnia says. “This is something I have heard distinctly from customers over the years — that they don’t want to go back to the dog days of being locked into a processer architecture or an operating system and having to be held hostage to a vendor and pricing around servers or what have you.”
I asked Behnia what some of the important cloud management practices are to consider. He offered his advice around planning, automation and control, and monitoring for performance, availability and capacity.
“Planning is very important, especially for private clouds. You just cannot create a cloud and say ‘here it is.’ The cloud is only as good as the applications and the work flows that live in it,” Behnia says. “Start by identifying what applications you have, what the current environment looks like, where the most appropriate areas are where you could virtualize those servers and plan how they would be utilized.” Behnia recommends using discovery technology that can look at what is out there, and then store that information in a configuration management database. From there, capacity management products can monitor utilization of those resources and make assessments around what candidates there are for virtualization and how to place these candidates close to each other. In other words, workload modeling.
The next step, Behnia says, is automation and control. “Define a set of standardized catalog entries or offerings, and create classes of services around them. For instance, you can imagine a bronze/silver/platinum model. Think of it like a store front for IT. Basically you’d have a product list so that your customers can view the prices and the SLAs. You can make the entire process self-service with an automated portal where customers can come in and order things, and there’s a check out process,” he says.
Behind the scenes, he recommends you connect the front-end process to a set of automated routines such as configuring, provisioning, de-provisioning, modifying capacity and so on. “Those automated routines are analogous to, say, ordering a book from Amazon.com. The processes might force you to ask, Have you fulfilled that request? How do you find where that virtual machine lives? How long does it live? How do you interface it with your change process, asset process, and all the other processes that need to be adhered to within an enterprise cloud? And, have you actually provisioned that service?”
Last but not least, as you deploy these services, Behnia says there are concerns about effectively monitoring them for performance, availability and continuous capacity so that you can anticipate where you might need to buy additional hardware or rebalance the workloads. This process, which BMC calls Service Assurance, means being able to take user transaction performance, system and virtualization performance, and predictive analytics and be able to understand performance and health in a private cloud environment.
I asked Behnia what customers should look for in a service level agreement, whether they are deploying applications in a public cloud or a private cloud. “Regardless of who is providing the service, the focus should be on business metrics and the things the business cares about and not the ‘five nines’ and the traditional IT metrics around infrastructure up-time,” he says. “I think this is where people struggle a bit with some of the public cloud vendors, because the SLAs will vary depending on what the application is.”
He gives the example of a financial services company. “This company cares about user response time when the traders are using a SaaS-based solution,” he says. “Some of this will entail network related performance that the provider has no control over. That’s why they are reluctant to provide end-to-end SLAs.” Behnia says an approach he has seen is for the IT organization to establish the end-to-end SLA to the service provider — such as Amazon, BlackBerry or whoever the network provider is — as the underpinning contract, which is a concept that ITIL has had for a long time and is a best practice around SLAs. Overlay and underpinned contracts are ways that you can roll up these SLAs to a point that you can have a dialog with the business in terms they can understand, which is much more around service availability and user transaction response times and less about infrastructure metrics such as uptime and things that are just plain too generic.
He concludes, “I do see the cloud providers becoming more aware of the issues around SLAs. I don’t’ know if the they can fundamentally address end-to-end SLAs, which is what customers ultimately care about. I think IT needs to take a much more active role in setting up what the end-to-end SLAs are and then evaluate vendors around the underpinning contracts and the overlays they would provide.”
For information on this topic, visit BMC Software’s Web site on cloud computing.
