SMBs face multiple decisions around migrating to public cloud, but there are some definite advantages that were covered in part one of this two-part series, including freeing up admins' time, enabling data center expansions, and easing security and availability concerns. Some enterprises, however, should stick with an internally managed vSphere infrastructure. The following factors may make your decision to stay with vSphere an easier one:
Paranoid security teams
No company wants to hear that its data has been compromised, but some companies cannot afford to take the risk. In such cases, IT infrastructures may need to stay in a data center, controlled by internal IT teams. This requires proper security using internal resources, but it also means having full control and awareness of the environment's security. Companies dealing with government regulation must consider security, which could prohibit migration to a public cloud.
Resource and allocation control
Very few public cloud providers will allow the level of access necessary for a single tenant to do extreme fine-tuning of the environment the way you can with VMware virtualization. A private vSphere implementation also eliminates the "noisy neighbor" phenomenon, where one virtual machine (VM) on a shared resource stack uses more resources than it needs to, inadvertently causing a resource strain on other VMs in the environment. This phenomenon could be caused by something as simple as flooding the network with traffic or as complex as CPU-scheduling within the hypervisor.
Some enterprises should stick with an internally managed vSphere infrastructure.
Support for bandwidth-intensive applications
When moving server applications away from end users, or when separating servers between two sites in a hybrid implementation, be sure the bandwidth between sites can support the traffic needs for this and all other applications. Placing the application server and the database server of a three-tier application in two different sites would probably not be a good decision because of the high amount of traffic between them. If end users have an application with a large amount of traffic flowing into a back-end server, then the server should be kept local to the desktops -- unless it is a virtual desktop and can migrate to the cloud as well.
The history of fear and politics in IT
While not technical reasons, the fear and politics surrounding an important decision such as moving to the public cloud are just as meaningful as bandwidth considerations, for example. Internal IT resources could fear losing their jobs or responsibilities; other negatives include the perception that public cloud can increase security risks.
Understanding the dollars and cents
Cost is the one great variable in the debate of migrating to public cloud or maintaining an on-premises vSphere VMware virtualization. Depending on the cost model, a purely internal implementation can be cheaper than a public cloud model. When deploying infrastructure internally, SMBs can design and implement components to suit their specific needs and individual applications. The level of redundancy inherent in a public cloud may not be necessary for all applications, and that can reduce the system implementation costs.
Overall, cost is neither a benefit of nor a detriment to moving to public cloud. Each public cloud provider must be evaluated individually on this point, taking into consideration capital expenditure, operational expenditures and "soft" costs such as employee time and efficiency. The best-case scenario would be that the cost of a public cloud environment would be equal to the cost of purchasing, building and operating an internal vSphere infrastructure. This allows SMBs to decide purely based on the technological factors above.
If you're undecided on whether to migrate to a public cloud or operate an internal vSphere infrastructure, use the four points below to see how your IT needs will be addressed with each option:
1. Understand the public cloud platform. How does the cloud provider secure the environment, both from the outside and between tenants? How does it prevent noisy-neighbor issues?
2. Consider a hybrid cloud model when concerned about ultra-sensitive data. Less mission-critical data can easily move to a public cloud platform, while critical data should remain hosted on an internally managed platform.
3. Negotiate realistic service-level agreements (SLA) with your public cloud provider. The cost of lost availability and/or data is not equal from one company to the next. It is imperative that each SMB negotiate its own SLA.
4. When nervous about running a full production environment in the public cloud, an SMB can test the waters by first moving less-critical resources such as disaster recovery (DR) there. Once comfortable with this approach, SMBs can flip DR to the internal data center and move production to the public cloud.
Disclaimer: The author currently works for a company providing on-site and colocation infrastructure services and has recently launched a public cloud platform.