Global Column | What is the cost of making the cloud more resilient?

Cloud service providers separately offer users the ability to increase resilience. You can increase your resilience by investing only the cost, but it is not easy to know the appropriate cost. Meanwhile, consulting firm Uptime Institute has published an interesting study, stating that increasing resilience to basic services without protection could cost up to 111%.
ⓒ Getty Images Bank

Uptime Labs’ ‘Public Cloud Cost vs. Resilience: Stateless Applications’ reportAccording to the report, by adding costs, businesses can recover their infrastructure from an outage more quickly and receive better compensation than the contract level stated in the SLA. This means that the reliability of the service can be improved to some extent.

Uptime Labs analyzed three scenarios for increasing resilience for testing. The website used was a simple WordPress-based site that maximized the request volume and confirmed that it responds within 3 seconds. We created a separate Python simulation here to see the cost impact of varying bandwidth and virtual machine requirements.

The analysis was conducted based on AWS, and the Uptime Research Institute introduced, “Since other public cloud services have similar pricing models, services, and architectural principles, the basic analysis results of this report can be applied to other cloud service providers as well.”

The report analyzed the resilience of WordPress apps by dividing them into ▲backing up the VM hosting the app to the same Availability Zone ▲backing up the VM to different Availability Zones in the same region ▲backing up the backup to different regions.

An availability zone of a cloud service refers to a virtual data center, and when the availability zones are gathered in the same geographic location, it becomes a region. “A single resource, such as a VM, can sometimes go into an unresponsive state,” the report said. Also, if all of the Availability Zones go down, many resources can go into an unresponsive state. “It’s rare for an entire region to go down, but once it does, it results in multiple Availability Zones going down.”

The cost of the basic service without protection, which is the core of the report, was priced at the cost of using the VM plus the cost of outbound bandwidth, for a total of $217.38 per month. If there is no support to back up apps in the event of a VM failure, the recovery time will ultimately depend on how quickly the customer replaces the failed app. “AWS’s data control plane is designed to achieve 99.95% availability,” Uptime Labs explains. As a result of testing by Uptime Labs, the compensation amount was estimated to be 29% of the monthly application cost in the event of an outage of 1.5 days or more in AWS.

Backup to the same Availability Zone

If you use a load balancer and back up your VMs as separate active VMs to the same Availability Zone, even if a VM fails, the downtime is zero, resulting in an equally implied availability of 99.95%. Compensation for downtime lasting more than 1.5 days is increased to 44% of the monthly cost. These architectures also cost more because they require additional VMs and load balancers. That means you need $311, which is 43% higher than the base rate.

Backup to 2 Availability Zones in the Same Region

Backing up the VMs as separate active VMs in different Availability Zones within the same region costs $311 per month. There is no additional cost to having a second VM in a different region, but implicit availability improves to 99.99%. The recovery time is zero and the compensation rate is also 44%, which is the same as in the aforementioned scenario.

Backup to another region

According to the report, installing the app in two different regions and hosting the two active instances in different Availability Zones in each region was the most resilient method.

Uptime Labs had four active virtual machines hosting apps, two in each region, with virtual load balancers in each region balancing traffic between VMs in different Availability Zones. According to the report, “Load balancers provide resiliency and simple balancing in the event of an outage in VMs or Availability Zones. Externally, the failure was either not visible to the end user or had to be managed by the end user’s device.”

Traffic to these virtual load balancers is sent by the Domain Name System (DNS). DNS can be configured to select a better load balancer based on physical proximity, delays in the path to the load balancer, weighting policy, etc. DNS checks its health to detect when a load balancer becomes unavailable and can even direct traffic to another load balancer.

This approach, which is the most resilient, also has its drawbacks. According to the report, “Using DNS can be problematic as it creates a record of the application IP address on the user’s device accessing the web application. If this address becomes unavailable, the user device will not be able to access the application until the application updates its local cache with the IP address from the DNS system.” Therefore, the end user may not be able to use the service without a commitment in the event of a failure.

In the last scenario, implicit availability rises to 99.9999% and costs 111% above the base rate to $457.80. According to the report, if one of the regions goes down, the load balancer and two VMs will be unavailable, so if the downtime lasts more than 1.5 days, customers will be compensated for 62% of the cost of the service.

Uptime Labs said that cloud service providers “in the availability zone, basically provide a lot of features to increase resiliency, and for a small additional cost, you can achieve quite high availability.” “It is also important to be aware that a more resilient architecture may not guarantee a meaningful level of availability or downtime compensation,” he explains.
[email protected]

Source: ITWorld Korea by

*The article has been translated based on the content of ITWorld Korea by If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!

*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.

*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!