Many IT professionals believe private cloud storage is merely another name for their current infrastructure and that public cloud storage is the same thing -- just implemented in someone's public data center. Vendors perpetuating these ideas typically lack cloud storage infrastructure technologies or are attempting to buy time as they catch up to the rest of the market.
A traditional IT infrastructure separates a data center into silos of functionality -- applications, compute, networking, storage and facilities -- and expertise. A siloed infrastructure is very good for slow growth, non-dynamic, predictable workloads. This type of infrastructure places all the responsibility for forecasting compute, networking and storage requirements -- such as throughput, IOPS, capacity, network bandwidth and flows -- on the buyer for the life of that infrastructure.
Changes tend to be slow and costly, so the risk often belongs to the buyer. To meet user requests, most IT professionals err on the side of caution by overestimating and overprovisioning their storage infrastructure. This results in higher costs and doesn't prevent unpredictable demands from exceeding the capabilities of that infrastructure because it is inelastic and not easily changed.
A private or public cloud storage infrastructure flips that script with options designed to have elastic compute, networking and storage capabilities. Elasticity, scalability, flexibility and low cost are the defining characteristics of a cloud infrastructure. However, there are notable differences between public and private cloud storage that clarify when and why a private cloud infrastructure makes more sense.
Public cloud storage infrastructure pros and cons
A public cloud infrastructure is multi-tenanted and shared. To any given user, it appears to be unlimited because they can expand their resources on-demand for unexpected workloads and shrink back when demand decreases. In reality, those resources are oversubscribed so that the service provider can make a reasonable profit. This is conceptually similar to oversubscribing a virtualized server. It is statistically unlikely that all of the tenants or users will require infrastructure resources at the exact same time; it does happen, but rarely. Tenants are protected by service-level agreements that provide some payment relief should they not get the service for which they are paying.
However, there are several well-publicized negatives to a public cloud storage infrastructure.
- Security control. The public cloud is quite secure; the issue is one of control. In high-compliance, regulated industries, such as healthcare, insurance, financial services and government, compliance responsibility cannot be outsourced. These organizations can use public cloud infrastructures, but they are still responsible for ensuring compliance. Since failure by a service provider does not absolve them of their responsibilities, few companies like to give up their ability to manage and control data security.
- The inability to fine-tune application workload performance directly. There are ways to manage the workload to a point, just not to the same degree as if the infrastructure was all in-house or private.
- Long-term cost. Short-term costs for a public cloud tend to be lower than investing in a private cloud, but long-term costs can be higher. Administrators complain about the hidden or unexpected costs of reading or moving data; however, many organizations prefer managing their business and business applications rather than their infrastructure. There are significant cost savings that can be attributed to not having a data center or having a smaller data center.
- Data stickiness. This negative often flies below the radar. Once tenant/user data is in the cloud, it frequently becomes difficult and inconvenient to move it out or change cloud service providers. As the amount of data increases in the public cloud, so does the difficulty of changing providers or pulling data back to a private cloud.
- The future. What happens when a public cloud provider becomes so oversubscribed that service is no longer acceptable? Worse, what happens if they become insolvent, like Nirvanix? Then, tenants must scramble.
Is private cloud a better choice?
A private cloud infrastructure uses the same technologies as a public cloud:
- Hypervisors or containers for software-defined compute.
- Open Flow switches, network functions, virtualization and application-controlled network controllers for software-defined networking.
- Software-based storage in hypervisor kernels or running on virtual machines or appliances for defined storage.
These private cloud infrastructures enable local security and performance control. There are no issues about provider solvency, oversubscription or data stickiness. Costs are fairly predictable and lower over the long haul. If security/performance control, compliance management and lower long-term cost are key concerns, a private cloud infrastructure could be the way to go.
Private cloud infrastructures still place the risk on the buyer of the infrastructure, though technology has reduced that risk. Elasticity is limited to the physical resources available. Tech refresh issues may still exist, depending on the vendor, as does vendor lock-in.
As with all technologies, there is no perfect answer. Which type of cloud storage infrastructure works best for a given organization depends on their specific requirements.
Justify a private cloud migration before making the switch
Private cloud vs. public cloud for cold storage
Challenges when using a cloud architecture with local storage