Reviewing the Bidding
An axiom in the computer storage world dictates that if you don’t have three copies of mission-critical data, eventually you won’t have any data.
Two copies generally are stored on locally redundant disk media, allowing data to be written without interruption when one disk inevitably fails. A third copy can be on cold backup storage, or ideally online in a remote region, preventing permanent loss of service in the low but possible chance of a site failure.
Disks fail in the cloud. Sites fail in the cloud. You need three copies of your mission-critical data written by your EC2 workload.
Two local copies are already accounted for with standard EBS storage backing EC2 instance: AWS’s platform automatically mirrors writes between independent disks in the same availability zone.
The third copy could be a redundant disk in a separate availability zone in the same region, but as I’ve argued repeatedly that testing the viability of this third disk is essentially impossible without powering down the primary availability zone to confirm your application can continue to run. An untested disaster recovery plan — like any software that is untested — is doomed to failure. DR cannot fail.
Backups are easy, but for use as disaster recovery are as dubious as availability zone redundancy for the same reason: there is no easy and automated way to test restoration on a regular basis.
Replicating volume snapshots across regions, and attaching those replicated snapshots to backup EC2 instances, is the most robust DR mechanism in the cloud and offers the quickest recovery time over restoring from backups. Most importantly, it is easily and automatically testable: just power on the backup EC2 instances regularly and connect to the backup application to confirm its recovery, for example by issuing a test SQL query to a database. You know the remote region will be accessible in case the primary is down: AWS has historically suffered brief site outages; none affected other sites.
There are many solutions on the market to automate the replication and failover of EC2 instances across AWS regions. Most of the them require heavy operating costs of hosting a 24/7 dedicated EC2 instances to run the software, as well as a hefty license fee. Though AWS’s own offering is not licensed, it requires a proxy replication server for each instance essentially doubling the operating cost of each EC2 instance. And all testing in their product is manual, which on your busy schedule probably ranks somewhere below cleaning out the coffee machine.
Our product Thunder for EC2 automates the provisioning of duplicate EC2 instances across regions, the realtime replication of snapshots across regions, and deep automated testing of backup instances to give you the most peace of mind that your backup infrastructure will start seamlessly in the one-in-a-million chance of a site disaster at the region hosting your production infrastructure.
The first and only enterprise solution implemented as a Lambda function, Thunder for EC2’s operating cost is essentially zero. Because snapshotting data only occurs infrequently throughout the day, the function only spends execution cycles when there is actually work to do, in contrast to SaaS software hosted on an EC2 instance which sits idle 99% of the day while you pay Amazon for the wasted cycles.
A dividend of using Lambda function means all supporting infrastructure is 100% cloud based: since we cannot host a web server in a function we dispose of the need for a custom UI, instead we use CloudFormation and CloudFront. Instead of using insecure log4j, we use CloudWatch Logs. Instead of using clunky cron, we use Eventbridge. Testing scripts are also managed through Lambda, and connect securely to backup instances through Security Groups on the private network. All deployment and ongoing management is through cloud services in the secure AWS console. This means or solution is lightweight, transparent, and has a minimal attack surface.
You need disaster recovery protection for your mission critical workload even in the cloud. In the past you had to choose among cobbling it together yourself (too much time), purchasing a legacy solution (too much money), or doing nothing (too risky). With Thunder for EC2 you can be up and running in less than hour, pay less per year than others charge per month, and build the most robust, verifiable, secure, and cost-effective disaster recovery plan for your business. Need proof? Check out the product in action in our account: no cost and no risk to you, it’s running on our credit card. Or to schedule a one-on-one demo reach out to us at info@thundertech.io
Originally published at https://www.linkedin.com.
