Managing cloud infrastructure and services is similar to traditional network management - only bigger, badder and more complex.
Where once you had to deal with maybe one or two strategic outsourcers, in the cloud world you're more likely contending with a dozen or more cloud service providers, be they software-as-a-service or infrastructure.
Read the other three parts to the Enterprise Cloud Services series.
Managing cloud infrastructure and services is similar to traditional network management - only bigger, badder and more complex.
Where once you had to deal with maybe one or two strategic outsourcers, in the cloud world you're more likely contending with a dozen or more cloud service providers, be they software-as-a-service or infrastructure.
Read the other three parts to the Enterprise Cloud Services series.
Where application workloads once moved over private links inside your data center, now they're flitting across the Internet.
Where server and storage capacity once fell to IT exclusively, now anybody can grab the resources they need, as quickly as they can pull up and fill out a Web form and enter credit card numbers.
So how are enterprise IT managers supposed to handle the supersized management challenges the cloud throws their way? Here's some advice for managing the cloud.
Have consistent data models
Sounds simple, but don't be fooled, says Beth Cohen, senior architect and consultant at Cloud Technology Partners, a cloud consulting firm.
Most companies have standard terminology in data records and databases to which cloud applications should adhere. This can be as basic as storing data with a standard ID number and using the same naming convention across CRM instances.
This is easy enough to control when IT is guiding the purchasing and the deployment, but what happens when the marketing department turns to Saleforce.com for its CRM needs, as does sales, but in a different project?
Business people bringing in applications via the SaaS model aren't necessarily going to be thinking on that level. And IT has got to get out in front of this issue, Cohen says.
"As long as the data models match when you want to orchestrate with other applications, either elsewhere in the cloud or internal to the enterprise, the integration process will be that much easier. And note, that is 'when' you want to do this, not 'if,' because this will be happening," Cohen says.
Integration, she adds, is a real struggle point. "It's not unsolvable; it's a technology problem. But IT had to be aware of it."
Plan for data integration
With integration of one sort or another all but inevitable as enterprise cloud use evolves, the smart IT department should be taking a lead on qualifying cloud providers with this tricky management issue in mind, Cohen says.
That could prove challenging, she says. "Most vendors haven't been too proactive about the integration piece. They're vertically focused and mostly concerned only about delivering their service and not about integrating with the 10 or 100 other applications a particular company might have."
At the American Hospital Association, in Chicago, no SaaS provider gets by IT's scrutiny - and IT does due diligence on all potential cloud service providers - without meeting a set of integration-related checklist items, says Karthik Chakkarapani, IT director, Technology Solutions & Operations.
Knowing how a potential provider will integrate with current and future SaaS applications, how it will work with the organization's hybrid cloud-based single sign-on (SSO) environment and how it provides database access are imperative, he says.
"The best way to interact with data is through Web services, so we ask what kinds of Web services they support, too," Chakkarapani adds.
Create a provider ecosystem
One of the biggest management headaches of having SaaS applications intertwined with each other and internal applications is coordinating updates and fixing issues with one that might affect the others, Chakkarapani says. This is an art, and where a strong, preferably IT Infrastructure Library-based internal service desk, is essential, he says.
The AHA requires coordination among about 30% of its SaaS-provided applications; the rest live in silos, Chakkarapani says. In one recent example, Symplified tested and validated that its SSO service worked with the latest social collaboration release from SaaS provider Socialtext prior to the AHA putting the upgrade into its production network, he says.
"When one vendor has an upgrade, both have to test before we can go into production. These issues are slowly starting to crop up, and the more and more we have, the more important it is that we have a good vendor ecosystem," he says.
Build a DevOps team
One of the hairiest infrastructure management issues for IT operations is actually not being able to manage resources at all. That scenario occurs when developers go around IT and grab resources in the cloud rather than wait on traditional internal provisioning processes.
Creating a DevOps team that can provide the rapid provisioning and super smart configurations required of today's most agile, cloud-oriented developers is one of the best ways to circumvent this problem, says Rachel Chalmers, an analyst with The 451 Group. This means IT must embrace a change in mindset - to one of a service provider - and a new toolset.
On the later point, Chalmers encourages DevOps teams to use cloud infrastructure automation tools from companies such as Opscode (see "10 cloud management companies to watch") or Puppet Labs.
Go for drag-and-drop simplicity
Being able to capitalize on the use of a fully dynamic private or hybrid cloud infrastructure requires a management tool that lets you do things like reduce cycle times, provide better automation, get a handle on resource consumption for chargeback purposes, ensure adherence to security standards and, of course, quickly and easily spin up new environments and scale resources.
This means adding a cloud management layer on top of what a company already has in place for virtualization management, says Dhiraj Pathak, director of PricewaterhouseCooper's CIO Advisory Services practice. "This is a distinct layer of capability, one that allows for the efficient management of these virtualized resources. It's still in its early days, with some parts of the layer maturing while others are yet to fully form," he says.
Building such an overlay that would allow for more automation and smarter resource use was the objective for Roundarch, a digital design firm in Chicago, when it went looking for an enterprise cloud management tool. Plus, says Geoff Cubitt, president and CTO, the company wanted to be able to let users manage things on their own, thus reducing the strain on the IT team and providing much-needed flexibility across hypervisor environments.
The ability to create, then drag and drop, reusable images and templates from one environment - testing to staging servers, for example -- to another is key when dealing with the cloud, Cubitt says.
He notes that Roundarch has met all of its cloud management objectives with Abiquo enterprise cloud management software. (Network World last spring recognized Abiquo as one of 15 cloud companies to watch for 2010.)
"Drag-and-drop image build lets us do things like spin up a standard environment - maybe that's the OS, the application server and a database hardened with our security standards and configurations that we like - and copy it to where we need it. This means I don't have to have my best systems administrators setting up these environments. Anybody can do this; it's so simple, but it gives us better leverage on our resources and speeds up cycle times while letting IT be more responsive," Cubitt says.
Account for multi-hypervisors
Look for an enterprise cloud management tool that will support multiple hypervisor - even if you're only using one today, experts advise. Lots of companies have built their virtual data centers around VMware but will increasingly look to bring in other hypervisors to drive costs down and gain more flexible provisioning options as they migrate to the cloud.
At Roundarch, for example, having the Abiquo management software facilitated the company's move away from VMware/Red Hat Linux to a lower-cost Xen OS environment, Cubitt says. Abiquo, which supports multiple hypervisors, allows Roundarch IT administrators to port images created in the VMware/Linux environment to the Xen world.
"That let us build out from our existing infrastructure and lets us leverage both internal and external clouds, being about to burst into public resources as appropriate. We wanted the flexibility to be able to manage across boundaries, and we got it," Cubitt says.
Getting the right management tool, PwC's Pathak agrees, is critical for a successful cloud deployment.
Look for help on cost management
The use of cloud resources, especially when business groups are making some of these decisions, seriously complicates the ability to capture and manage the total cost of IT, says Phil Garland, a partner in PwC's CIO Advisory Services practice.
"We find integrating of the cloud fulfillment, particularly if resources are coming from the public cloud to the standard procurement process within the enterprise, is an area where people get surprised. It's a difficult connection to establish, but it's so important to be able to effectively track consumption of resources from the cloud and the cost of consuming those resources," Pathak says.
Toward that end, he says, look for cloud management tools that incorporate financial engineering aspects of cloud services. That's a major differentiator starting to emerge among tool makers, with some enabling mapping against specified service-level agreements, he says.
And if introducing cloud services internally, he adds, IT needs to develop a consistent cost model. "Transparency is a big expectation users have around the cloud, so you don't want to be costing out every service on an ad hoc basis."
Leave no discipline untouched
Overall, the challenges of cloud management are similar to traditional management, but bigger and "badder." All the same disciplines enterprise IT organizations have applied to their legacy environments have a place here, too.
This includes the application, network and systems management realms, as well as overarching programs like governance, policy orchestration and SLA management.
So, power up on management capabilities before plowing into the cloud.