Any technology in its early stages can have destructive preconceptions; cloud computing is no different. And one...
of the most destructive cloud assumptions is that the lowest price wins.
Companies adopting a cloud model make the mistake of opting for a vendor with the lowest price for their chosen type of cloud service. This is a big mistake; an enterprise needs to align its mission with the best cloud service provider, taking into account three aspects of the provider's offering: software, application availability and data requirements. The following are three questions related to these aspects that all enterprises should ask before choosing a cloud service or provider.
Where does the cloud software come from?
The primary concern an enterprise should have of a prospective cloud consumer is where the cloud software comes from. For an enterprise customer, all cloud services must generate some type of Software as a Service (SaaS) for end users to consume. So, the cloud platform resides at the bottom of the stack; application software resides at the top.
Cloud services such as SaaS offer applications directly to users, but all other cloud services require end users build upward from the cloud platform to the user interface. Cloud consumers must cover the cost of the necessary software layers, including licensing and support.
Companies that have already written or plan to write their own software likely will discover that a Platform as a Service (PaaS) provider that uses the same OS and middleware offers a more cost-effective strategy. PaaS reduces not only hardware costs but also platform software costs for both acquisition and support. If PaaS does not meet the enterprise's needs, Infrastructure as a Service (IaaS) likely will.
One of the most destructive cloud assumptions is that the lowest price wins.
If your enterprise plans to use third-party software, you may want to find a cloud provider that hosts some or all applications and offers SaaS directly or through a partner. Amazon Web Services and Rackspace probably host the majority of third-party applications available in the cloud, but check with your software providers to see if and where they host their applications. Even if you plan to supplement SaaS with custom software or Web-based apps, having a single provider for both SaaS and in-house (contributed) software (running on IaaS or PaaS) will simplify application integration and deployment.
Where SaaS applications are going to be supplemented by user-supplied software, it's usually easier to integrate the two applications if the SaaS provider hosts all associated software. This is another reason to closely examine a cloud provider's SaaS offerings to determine if it simply provides the platform or its own cloud services. If the cloud provider also offers IaaS or PaaS that's compatible with the enterprise's internal applications, it could be the ideal package, because this facilitates management and integration.
How mission-critical are your cloud applications?
The more mission-critical the enterprise applications are, the more resilient the cloud must be.
Pilot cloud projects, test and development, as well as some Web-related activity may not have a critical availability requirement and can often be aligned with cloud services lacking high-availability features, which are often less expensive.
However, cloud provider availability can be difficult to measure, and if you need higher availability for your more critical applications, it can mean it may also be difficult to write a meaningful service-level agreement (SLA) for cloud services. But there are two steps you can take to match cloud provider availability with your needs:
- Look for availability options such as multiple-service/geographic zones that can protect you from power outages. If no such options are available, you may be vulnerable to local power outages, which can be a source of cloud failures.
- Ask whether you can attach the cloud service to the enterprise's VPN or virtual LAN (VLAN). Internet connections to the cloud are typically the weakest link in terms of availability, but VPNs can be associated with specific SLA guarantees of availability. VPN or VLAN connections often offer higher performance.
What are your enterprise data needs and where does cloud computing fit?
If applications have modest data needs, then cloud data pricing and performance won't likely make much difference. But as data use grows, enterprises either end up paying high cloud data storage charges or storing data on-premises and linking that data to the cloud. The latter can create serious performance bottlenecks if access connections are slow.
Cloud providers differ in their support for storage and data models. Most offer block and object store, while some also offer data-distributed cloud storage models, such as Hadoop and relational database management systems. Applications with higher data use must be audited against available data models to ensure they don't incur a higher cost than anticipated.
What is the long-term value of the virtualization model?
For enterprises looking to move forward with their cloud computing project, the most important question regarding a cloud mission may involve the longer-term value of the entire virtualization model. Current cloud applications tend to be drawn from server consolidation and Web-related activity, while monolithic applications written for independent servers dominate.
In the future, cloud computing will drive both software providers and cloud providers to a "component-hosting" model similar to service-oriented architecture (SOA). This model will increase the value of PaaS and clouds based on operating systems such as Linux or Solaris. These cloud types should be considered as strong contenders for buyers with an eye to the future.
Tom Nolle is president of CIMI Corp., a strategic consulting firm specializing in telecommunications and data communications since 1982.