In an ideal world, customers would be able to fully take advantage of the benefits of hybrid cloud by rationally matching infrastructure parameters -cost, performance, reliability, availability, security, regulatory compliance, scalability- with the requirements and dependencies of each application.
In reality, customers are running into the following 3 challenges:
Customers are typically faced with so many choices of private and public server, network, storage and security components, that it becomes virtually impossible for them to identify the optimal deployment target. Therefore, many enterprises lean toward an “insurance policy” that includes significant over-provisioning combined with the unnecessary use of tier 1 server, network and storage infrastructure. Based on EMA research, this is an insurance policy that adds approximately 25% to the cost of the average application environment.
Vendors such as Cirba, Cloudability, CloudPhysics, CliQr, Turbonomic and RightScale offer software to automatically and continuously evaluate workloads based on corporate policies and match them with the deployment targets that are the closest fit. Customers can then choose, on a per application basis, how aggressively they want to cost-optimize their environments, versus building in some slack to minimize risk.
“We can save a lot of money by leveraging Amazon Web Services, Microsoft Azure or IBM Softlayer for hosting our new web application.” Whether this statement is true or false depends on how the characteristics of the application workload correspond with the consumption based pricing model of each cloud provider. To optimally match applications to deployment targets, their resource consumption profile has to be determined and compared to the target’s pricing model. In addition, there is a risk that limitations in terms of network or storage performance negatively affecting the end user experience or even application availability, during peak times. This can lead to costly SLA violations.
The above mentioned solutions by Cirba, Cloudability, CloudPhysics, CliQr, Turbonomic and RightScale are able to closely predict the cost and performance of an application, hosted on a specific cloud.
Moving applications from the data center to a public or private cloud or from one cloud to another is no trivial task and requires significant planning, re-engineering and testing. Due to differences in servers, network and storage, enterprise applications are typically locked into the infrastructure environment they were originally deployed to.
Workload Mobility in the Hybrid Cloud
To restore application portability, there are numerous vendors that translate server, network, storage and security requirements between cloud and datacenter technologies: Zero Stack, Platform 9, Flexiant, Stratoscale, CloudVelox, VeloStrata, ScalR and Incontinuum are examples for this vendor category.
To truly benefit from hybrid cloud, it is key to establish a central strategy that for application workload placement. This strategy must enable business owners to define the level of risk they are willing to accept in return for cost savings and deployment speed. It must also enable IT to make a rational and holistic decision in terms of the level of lock-in they are willing to accept.
Next Tuesday, I will talk about how Machine Learning and Cognitive Computing can bring the business alignment of hybrid cloud to a whole new level.