Most enterprises see public cloud computing as a cooperative technology in the data center; they'll cloud-source...
some applications, but not all of them. When building a hybrid cloud, the trick is knowing which apps to cloud-source and when.
Mapping the right applications to an enterprise's hybrid cloud vision will almost certainly determine how much value it can draw from the public cloud. The majority of enterprise IT applications fit into one of the following categories, each of which has its own hybridization issues:
- Transactional Web applications: serve Web-connected users and provide entry, customer support or other transaction-oriented activities.
- Private transactional applications: provide worker access to company applications, but do not provide general access to the public or customers.
- SOA-distributed applications: built on a service-oriented architecture componentization model where you can run a portion of the application in a distributed way.
- Data mining applications: access historical data to glean business intelligence (BI).
Transactional Web applications not only support customers or suppliers, but also work with Web front-end applications in the data center and can be accessed by remote workers. Transactional Web cloud models are the most commonly deployed application models in the data center.
The current industry practice is to use public cloud services to host these applications. A VPN link is then used to pass transactions from the Web server to an application server located within the company's data center.
Enterprise data remains in the data center, so information security and data storage costs are lower. The only significant security and compliance issue you may encounter involves securing the Web host-to-application server connection; offloading Web hosting can improve data center resource usage and costs. If you're looking into hybrid cloud, this is a good place to start.
Private transactional is more a problematic application model because it raises the question of where data is stored. Public cloud data storage fees vary considerably; some can be so high that charges for a single month would equal the cost of a disk drive of the same capacity. In enterprises where storage charges are a factor, it's best to store data in the data center and then access it using a remote SQL Server query.
The private transactional model is valuable in scenarios where a worker transaction requires users to access a large amount of data. For security reasons, most enterprises in this situation look for cloud providers who offer provisioned VPN service integration to their cloud computing product. A provisioned VPN lets enterprises obtain a meaningful application service-level agreement (SLA) that the Internet cannot provide.
SOA-distributed applications can increase data storage costs if the application components hosted in the cloud require access to the enterprise data warehouse; remote SQL Server queries can remedy this as well. But distributing application components via SOA can create additional security and performance challenges, particularly if you use the Internet to link enterprise users, enterprise-hosted SOA components and cloud-hosted components.
When building a hybrid cloud, the trick is knowing which apps to cloud-source and when.
In this instance, it's important to have a private VPN connection to the cloud via a network provider with a stringent SLA. Even when accessing the cloud through a VPN, you should review how application components are hosted to ensure that work isn't being duplicated between the data center and the cloud. Data duplication could cause delays and negatively affect worker productivity.
Out of all four of the enterprise IT applications discussed, data mining can create the highest data storage expenses. Despite this, BI and data mining applications are two of the most interesting candidates for to cloud-source because of their unpredictable nature. The option to allow cloud-based data mining or BI applications to access the data center via a remote SQL query logic is an option, but the volume of information that would need to move across a network connection could create performance problems.
The best approach to moving data mining applications to a hybrid cloud might be to create abstracted databases that summarize the large quantity of historical data enterprises collect. You then can store the summary databases in the cloud and access them via public cloud BI tools -- without racking up excessive charges. Additionally, there are minimal risks to storing detailed core data in the cloud.
ABOUT THE AUTHOR
Tom Nolle is president of CIMI Corporation, a strategic consulting firm specializing in telecommunications and data communications since 1982.