De EDB-blog
May 7, 2019

Moving enterprise infrastructure into the cloud is often seen as a noble pursuit. The rationale behind it is strong: Leveraging the distributed nature of cloud computing makes it possible to enjoy high availability, better scalability, and a reduced dependence on in-house hardware. Even better, cloud solutions often promise lower costs, providing processing power and resources at easily accountable flat rates.

When The Cloud Isn't Cost-Effective

In reality, however, moving operations into the cloud isn’t always the cost-saving panacea most enterprises expect it to be. Rather than providing around-the-clock enterprise readiness, cloud solutions can end up raising new questions and causing additional problems, turning what seems like a simple solution into a potential long-term headache. Before moving enterprise operations into the cloud, it’s crucial to consider the particular needs of your organization and weigh them against more traditional options like running on bare metal.

One major issue often encountered by enterprises moving into the cloud is the expensive disconnect between the anticipated costs of service and the actual costs. For SLAs with popular cloud services like Amazon Web Services (AWS) and Microsoft Azure, organizations often believe they’re only paying for processing power they need. In reality, enterprises must also account for the cost of storage — and when dealing with databases, the costs of disk space alone can be highly prohibitive.

Similar cost discrepancies arise when considering performance. When an enterprise requires high IOPS capabilities for a high-performance database, for example, a platform like AWS or Azure will strictly enforce performance limits based on the capabilities you’re paying for. Compared to bare metal, where even a 15,000 IOPS operation can burst slightly higher and still be handled by your hardware, the cloud can cause immediate performance degradation with no burstability.

When Scaling Via The Cloud Isn't the Answer

These frustrating scenarios relate to a common solution promised by cloud providers: scalability. While it’s true that it’s easy to scale cloud operations based on the changing needs of an enterprise, in the vast majority of scenarios, organizations end up scaling up, not down. The promise of being able to scale back may seem like a positive long-term cost-saving possibility, but most enterprises end up needing more power and scale — which ultimately leads to a higher operating cost.

Considering the costly downsides associated with cloud providers, it’s important for an enterprise to fully understand the needs, desires, and potential costs of moving to the cloud. For example, a cloud provider may offer 99.9 availability through an SLA, but your organization might require three or four nines. Cloud operations may often be stable but are not immune to outages, disrupting access to your critical information for indeterminate amounts of time. Unlike backing up on bare metal systems, which leaves you with access to physical hardware should you need to recover from a disaster, cloud-based backup and restoration procedures can be thwarted by a server that simply can’t be reached.

If Not The Cloud, Then What?

In terms of value and performance, this is why bare metal continues to offer enterprise operations one of their best options. Even an outdated MacBook Pro can easily match the performance of an m5.xlarge AWS environment at 1/5th of the cost. Cloud solutions may seem an easy, popular choice, but can end up significantly increasing the cost of operations without a significant improvement in performance.

When accounting for disaster recovery, bare metal also provides immediate access to hardware that can fully restore operations. If a cloud region goes down, even if you’re paying for persistent storage, you’re at the mercy of the cloud provider to restore access to an instance that’s been powered down. This leaves you with two choices: wait until access is restored, or restore operations from your own bare metal backups. If redundancy is important, cloud operations that also require an investment in additional backup methods will end up costing more money in the long run.

Even though there are many reasons cloud deployments are worthwhile endeavors, such as moving an enterprise to a DevOps environment, it’s still important to compare the cost and performance of cloud providers to bare metal. Accounting for availability and performance, bare metal often provides significant performance advantages over cloud providers at a fraction of the cost.

Stephen Holt is an Executive Technical pre-sales specialist in the North East USA. He has over 20 years of experience working with enterprise databases ranging from development, deployment, and performance tuning, along with assorted customer facing roles. Stephen now concentrates on leveraging...