Editor's note: This article is part two of a three-part series on multicloud strategies. Click here to read part...
one and here to read part three.
Multicloud computing -- when organizations use a mix of cloud computing services from different vendors -- offers a number of advantages. But multicloud users must also be careful to avoid unnecessary complexity and costs. While most companies are long-term multicloud candidates, few will balance their applications and planning across all of their cloud choices.
The first step in effective multicloud planning is to pick the primary cloud model for your business.
If you've already committed to a specific cloud vendor, there is a good chance that provider -- or the one you use the most, if you use several -- is your primary provider for multicloud planning purposes. The type of service that your primary provider offers -- software as a service (SaaS), platform as a service (PaaS) or infrastructure as a service (IaaS) -- is the cloud model around which you should plan and integrate your multicloud environment.
If you haven't yet committed to a cloud provider, assess your applications to determine your primary cloud model. To do this, identify how the three cloud models mentioned above map to your applications and IT policies.
Another approach is to look at all your potential cloud applications and assign them to a cloud model. However, this can be difficult without some base cost projections. As a baseline, start with the market-leading provider for each class in your geography -- for example, Amazon for IaaS, Microsoft for PaaS and Salesforce for SaaS. For each application, ask whether the market leader could support the application and what the rough cost of doing so would be. It's best to look at SaaS first, then PaaS and then IaaS. SaaS and PaaS will generally be more restrictive, but will they displace more of your current costs.
When you've tallied all of the applications and costs for each cloud model, the one with the largest total cost should be your primary cloud model for multicloud planning.
If most of your applications aren't in the cloud, your primary cloud model is your own data center -- and part of your multicloud plan should include integration between the cloud and your data center.
The challenges of multicloud computing
Moving an application to a cloud may not be simple, but it's fairly well understood. You pick a provider that matches your application features, compare the cost to your internal costs and look at how you'll integrate any workflows between the cloud application and the rest of your IT processes.
Multicloud computing complicates this because you're choosing several providers. The choice of one might affect the costs of the others because of differences in integration costs. You also have to consider the cost of managing multiple clouds and doing problem isolation if you have an application issue. Management and integration are the toughest issues with multicloud computing -- but you also have to prepare for potential problems when assigning applications to the right cloud.
Matching applications to the right cloud is much like picking a provider. If you followed the sequence of matching features and benefits across cloud service types and then within them, you'll identify cloud provider choices that have distinct capabilities. You'll also identify how to use these capabilities. As new applications become candidates for the cloud, apply this matrix of features and benefits to ensure that you pick the service that best fits the application.
During this application-to-service mapping, watch for the essential apps most often linked to SaaS. Because SaaS requires little technical support and oversight, your business units may push applications into SaaS -- even when it isn't the most economical option, or when integration problems will arise. Make sure all applications undergo at least a feasibility review before a platform is chosen.
Technical management of the cloud is a step many users miss. It's rare that a company can shift its entire business to a cloud service all at once. Normally, each application is hosted individually, which means data flowing between applications requires proper connections.
Identifying the source of each problem is also a requirement. If every one of your cloud providers is independent, you'll have to use different techniques for each of them and introduce top-level problem determination to direct issues to the proper provider.
Tips to simplify multicloud management
To simplify the management of multiple cloud providers, make them look like a single provider. Cloud services have APIs that let users deploy and manage their applications, and also formats through which applications and data can be loaded and extracted. Look at your potential providers' APIs and formats and see whether they offer at least one compatible subset. That way, you can control all of your clouds with the same tools.
Even if the best cloud providers don't have compatible formats and management APIs, there's still hope. A cloud management tool, such as RightScale or Cloudify, can give you a standard set of interfaces and capabilities across your multicloud environment.
Organizations with strong in-house technical support may be able to use DevOps tools, such as Chef and Puppet, to deploy applications across different public cloud platforms. These tools can also support applications that stay in your data center.
Don't forget the issue of integrating workflows in and out of your clouds. Cloud providers offer some help here, and many third-party tools are available. If your applications are primarily from a vendor like IBM, Microsoft, Oracle or SAP, check with them first to see which tools they can provide. Ask your cloud providers for suggestions as well.
Whichever tools you pick to organize, deploy and connect your multicloud computing environment, be sure the vendors that provide them have a track record of following cloud and business trends. You don't want to be constrained in the future by your current cloud management tools.
Multicloud adoption soars in the enterprise
Multicloud model brings data portability challenges
Reducing risks through a multicloud deployment