With the average cost of downtime estimated at $8,850 per minute, businesses can’t afford to risk system failure. Full access to services and data anytime, anywhere is one of the main benefits of cloud computing.
By design, many of the core services with the public cloud and its underlying infrastructure are replicated across different geographic zones. This helps ensure the durability and availability of your data and services and protects against downtime. However, outages happen. To protect against costly downtime, many companies spread their services across multiple providers to reduce the chances of failure.
But is a multi-cloud strategy the only solution for ensuring high availability?
In February 2017, an engineer’s typo caused a major service disruption on Amazon S3 in its US East region. The outage impacted many companies that relied on S3, and specifically those that relied on S3 exclusively in the region.
Workloads impacted by the S3 disruption fell into two categories: those considered “not mission critical” and those that lacked sufficient architecture and chaos testing. Companies who lacked a robust architecture sufficient for testing felt the impact most acutely. In this instance, replicating files on another cloud provider could have mitigated the effects of the disruption. However, cross-cloud replication would also add more complexity, perhaps unnecessarily. Using a single cloud provider with cross-region replication is another solution.
Let’s explore the technical feasibility of using multiple cloud providers to achieve high availability in three scenarios:
Application Distribution: Teams will have to work to abstract away vendor-specific functionality if they want to achieve high availability for the same functionality across different cloud providers. This means that you will be limited to the features that are common to your selected platforms. At the individual service level, the differences between various cloud providers’ implementations can create a lot of extra work in the form of abstraction layers.
Containers: At the application level, due to IaaS implementation differences across providers, containers could serve as a viable abstraction. This approach would require running the same container orchestrator on multiple platforms and limiting the use of underlying functionality (or accessing underlying functionality through a common interface). While using containers to run the same application across providers may be technically possible, the implementation is far from practical, making it more prone to human error and potential outages down the road. The potential increase in errors may be caused by differences in how data is replicated and differences in the IaaS offerings themselves.
Security and Compliance: Managing security for any single deployment across multiple public clouds will not be easy. Serving up virtual networks, firewall rules, monitoring, logging, and identity and access management can be difficult and time-consuming. Ensuring compliance across multiple providers adds a whole new level of complexity, especially at the rate that cloud providers release updates. Additional tooling, processes, and training will be required to ensure cross-platform consistency.
Is multi-cloud a solution for high availability?
New tooling or processes should be added to solve problems, not side effects of other problems. Adding the tooling required to implement a multi-cloud deployment is solving a side effect of using multiple platforms to accomplish what could be done with a single platform.
The bottom line is this: Multi-cloud could theoretically solve certain high availability issues, but it’s more likely to add undue complexity. Instead, a better understanding of technology and implementing best practices should be your starting point before looking for a multi-cloud solution.
This post is excerpted from our new whitepaper, Separating Multi-Cloud Strategy from Hype: An Objective Analysis of Arguments in Favor of Multi-Cloud.
You will learn:
- The reality vs. hype of multi-cloud deployments
- How to achieve high availability while avoiding vendor lock-in
- The advantages of a best-fit technology approach
- The arguments that should be driving your multi-cloud strategy