Cloud Database Rewards, Risks & Tradeoffs

Considering a fully-managed cloud database? Consider the top rewards, risks, and trade-offs related to performance and cost.

What do you really gain – and give up– when moving to a fully managed cloud database?

Now that managed cloud database offerings have been “battle tested” in production for a decade, how is the reality matching up to the expectation? What can teams thinking of adopting a fully managed cloud database learn from teams who have years of experience working with this deployment model?

We’ve found that most teams are familiar with the admin/management aspects of  database as a service ( a.k.a “DBaaS”). But let’s zero in on the top risks, rewards and trade-offs related to two aspects that commonly catch teams off guard: performance and cost.

Bonus: Hear your peers’ firsthand cloud database experiences at ScyllaDB Summit on-demand

Cloud Database Performance

Performance Rewards

Using a cloud database makes it extremely easy for you to place your data close to your application and your end users. Most also support multiregion replication, which lets you deploy an always-on architecture with just a few clicks. This simplicity makes it feasible to run specialized use cases, such as “smart replication.” For example, think about a worldwide media streaming service where you have one catalog tailored to users living in Brazil and a different catalog for users in the United States. Or, consider a sports betting use case where you have users all around the globe and you need to ensure that as the game progresses, updated odds are rolled out to all users at the exact same time to “level the playing field” (this is the challenge that ZeroFlucs tackled very impressively).

The ease of scale is another potential performance-related reward. To reap this reward, be sure to test that your selected cloud database is resilient enough to sustain sudden traffic spikes. Most vendors let you quickly scale out your deployment, but beware of solutions that don’t let you transition between “tiers.” After all, you don’t want to find yourself in a situation where it’s Black Friday and your application can’t meet the demands.  Managed cloud database options like ScyllaDB Cloud let you add as many nodes as needed to satisfy any traffic surges that your business is fortunate enough to experience.

Performance Risks

One performance risk is the unpredictable cost of scale. Know your throughput requirements and how much growth you anticipate. If you’re running up to a few thousand operations per second, a pay-per-operations service model probably makes sense. But as you grow to tens of thousands of operations per second and beyond, it can become quite expensive. Many high-growth companies opt for pricing models that don’t charge by the number of operations you run, but rather charge for the infrastructure you choose.

Also, be mindful of potential hidden limits or quotas that your cloud database provider may impose. For example, DynamoDB limits item size to 400KB per item; the operation is simply refused if you try to exceed that. Moreover, your throughput could be throttled down if you try to pass your allowance, or if the vendor imposes a hard limit on the number of operations you can run on top of a single partition. Throttling severely increases latency, which may be unacceptable for real-time applications. If this is important to you, look for a cloud database model that doesn’t impose workload restrictions. With offerings that use infrastructure-based cost models, there aren’t artificial traffic limits; you can push as far as the underlying hardware can handle.

Performance Trade-offs

It’s crucial to remember that a fully-managed cloud database is fundamentally a business model. As your managed database vendor contributes to your growth, it also captures more revenue. Despite the ease of scalability, many vendors limit your scaling options to a specific range, potentially not providing the most performant infrastructure. For example, perhaps you have a real-time workload that reads a lot of cold data in such a way that I/O access is really important to you, but your vendor simply doesn’t support provisioning your database on top of NVMe (nonvolatile memory express) storage.

Having a third party responsible for all your core database tasks obviously simplifies maintenance and operations. However, if you encounter performance issues, your visibility into the problem could be reduced, limiting your troubleshooting capabilities. In such cases, close collaboration with your vendor becomes essential for identifying the root cause. If visibility and fast resolution matter to you, opt for cloud database solutions that offer comprehensive visibility into your database’s performance.

Cloud Database Costs

Cost Rewards

Adopting a cloud database eliminates the need for physical infrastructure and dedicated staff. You don’t have to invest in hardware or its maintenance because the infrastructure is provided and managed by the DBaaS provider. This shift results in significant cost savings, allowing you to allocate resources more effectively toward core operations, innovation and customer experience rather than spending on hardware procurement and management.

Furthermore, using managed cloud databases reduce staffing costs by transferring responsibilities such as DevOps and database administration to the vendor. This eliminates the need for a specialized in-house database team, enabling you to optimize your workforce and allocate relevant staff to more strategic initiatives.

There’s also a benefit in deployment flexibility, Leading providers typically offer two pricing models: pay-as-you-go and annual pricing. The pay-as-you-go model eliminates upfront capital requirements and allows for cost optimization by aligning expenses with actual database usage. This flexibility is particularly beneficial for startups or organizations with limited resources.

Most cloud database vendors offer a standard model where the customer’s database sits on the vendor’s cloud provider infrastructure. Alternatively, there’s a “bring your own account” model, where the database remains on your organization’s cloud provider infrastructure. This deployment is especially advantageous for enterprises with established relationships with their cloud providers, potentially leading to cost savings through pre-negotiated discounts. Additionally, by keeping the database resources on your existing infrastructure, you avoid dealing with additional security concerns. It also allows you to manage your database as you manage your other existing infrastructure.

Cost Risks

Although a cloud database offers scalability, the expense of scaling your database may not follow a straightforward or easily predictable pattern. Increased workload from applications can lead to unexpected spikes or sudden scaling needs, resulting in higher costs (as mentioned in the previous section). As traffic or data volume surges, the resource requirements for the database may significantly rise, leading to unforeseen expenses as you need to scale up. It is crucial to closely monitor and analyze the cost implications of scaling to avoid budget surprises.

Additionally, while many providers offer transparent pricing, there may still be hidden costs. These costs often arise from additional services or specific features not covered by the base pricing. For instance, specialized support or advanced features for specific use cases may incur extra charges. It is essential to carefully review the service-level agreements and pricing documentation provided by your cloud database provider to identify any potential hidden costs.

Here’s a real-life example: One of our customers recently moved over from another cloud database vendor. At that previous vendor, they encountered massive unexpected variable costs, primarily associated with network usage. This “cloud bill shock” resulted in some internal drama; some engineers were fired in the aftermath.

Understanding and accounting for these hidden, unanticipated costs is crucial for accurate budgeting and effective cost management. This ensures a comprehensive understanding of the total cost of ownership and enables more informed decisions on the most cost-effective approach for your organization. Given the high degree of vendor lock-in involved in most options, it’s worth thinking about this long and hard. You don’t want to be forced into a significant application rewrite because your solution isn’t sustainable from a cost perspective and doesn’t have any API-compatible paths out.

Cost Trade-offs

The first cost trade-off associated with using a cloud database involves limited cost optimizations. While managed cloud solutions offer some cost-saving features, they might limit your ability to optimize costs to the same extent as self-managed databases. Constraints imposed by the service provider may restrict actions like optimizing hardware configurations or performance tuning. They also provide standardized infrastructure that caters to a broad variety of use cases. On the one hand, this simplifies operations. On the other hand, one size does not fit all. It could limit your ability to implement highly customized cost-saving strategies by fine-tuning workload-specific parameters and caching strategies. The bottom line here: carefully evaluate these considerations to determine the impact on your cost optimization efforts.

The second trade-off pertains to cost comparisons and total cost of ownership. When comparing costs between vendor-managed and self-managed databases, conducting a total cost of ownership analysis is essential. Consider factors such as hardware, licenses, maintenance, personnel and other operational expenses associated with managing a database in-house and then compare these costs against ongoing subscription fees and additional expenses related to the cloud database solution. Evaluate the long-term financial impact of using fully-managed cloud database versus managing the database infrastructure in-house. Then, with this holistic view of the costs, decide what’s best given your organization’s specific requirements and budget considerations.

Additional Database Deployment Model Considerations

Although a database as a service (DBaaS) deployment will definitely shield you from many infrastructure and hardware decisions through your selection process, a fundamental understanding of the generic compute resources required by any database is important for identifying potential bottlenecks that may limit performance.

For an overview of the critical considerations and tradeoffs when selecting CPUs, memory, storage and networking for your distributed database infrastructure, see Chapter 7 of the free book, “Database Performance at Scale.”

After an introduction to the hardware that’s involved in every deployment model, whether you think about it or not, that book chapter shifts focus to different deployment options and their impact on performance. You’ll learn about the special considerations associated with cloud-hosted deployments, serverless, containerization and container orchestration technologies such as Kubernetes.

Access the complete “Database Performance at Scale” book free, courtesy of ScyllaDB.