Yesterday Amazon Web Services (AWS) had a service disruption in its S3 service in US-East-1 region. Approximately 148,213 websites got impacted. Good news – They were back up and running in around four hours! Bad News – They were down for around four hours!
My thought: Resistance is futile! People may have the urge to bash public clouds and AWS specifically and advocate not using cloud services. We have to let that knee-jerk reaction go. The biggest benefits of cloud architectures, the speed at which things can get done is too attractive a proposition to be thrown away because of a rare outage. If anything – cloud vendors will work even more harder to reduce such incidents. All the same, this is a reminder and an opportunity for enterprise technologies to make sure they are building redundancy into the application and data tiers of any of their IT services – whether on-prem or in the cloud.
For example, if most of these websites were built to store their production application data in multiple regions, and had the awareness in application to switch to a different region, their downtime could have been minimized. Yes this comes at an extra steady state cost of storing another copy in object storage at a remote region. But that’s the price you pay for redundancy and availability. After all, in your own data center, don’t you always use 2 firewalls, 2 network switches, clustered compute for key applications and redundant storage? Don’t you replicate data to a remote data center or colo to survive a site or regional disaster?
Design considerations for such data redundancies in cloud is vital to survive cloud outages, the same way you need data redundancies for internally hosted applications.
Here are a few ways that Actifio can help you survive such outages:
Use Case 1: Recovering On-Premises Production Applications to the Cloud
You can use an Actifio appliance to backup and retain a local copy of production application data on premises, to do instant recoveries for day-to-day operations. You can also use it for instant provisioning of Data as-a-service to DevOps teams, Security or Analytics teams. For recovery into major public cloud, Actifio replicates data to clouds in a few ways.
- As in Figure 1, Deploy an Actifio software appliance (we call it “Sky”) in cloud compute, and store data on the cloud’s block storage tier (EBS in AWS, Block in Azure). Replicate data from the on-premises Actifio appliance to a persistent target Sky appliance in the cloud (with encryption, compression, and optional deduplication). Use replicated data copy (which is on cloud block storage), to provision data as a service for not only on-demand DR, and DR tests, but also for Dev/Test in cloud.
- Use Actifio software on-premises to replicate data to cloud object storage (like S3 / S3 IAS in AWS or Azure Blob). See figure 2. At the time of DR or DR test, Actifio gives you the ability to access data rapidly, within minutes, from object storage to on-demand compute. You can also deploy a new Sky appliance, as needed, to make management of that data in a recovery state easier and faster.
Use Case 2: Inter-Region DR for Production Applications in the Cloud
Deploy Actifio Sky software in the same region where production enterprise apps are running. Use Actifio Sky to create a local backup copy in that region, and then replicate to Actifio Sky or object storage in another region of the same cloud vendor, e.g. from AWS US East to Canada Central. Thus with an outage in one region, you can bring up your applications in the second region quickly with Actifio in the remote region. This preserves independence from proprietary cloud infrastructure-based protection schemes, adding redundancy, while still working within a single preferred cloud vendor.
Use Case 3: Recovery of Production Applications to a Different Cloud Vendor
If you want extra peace of mind, and protection from multi-regional outages within a single cloud vendor s, you could consider using a second cloud vendor as your DR site. For example, if your production application is hosted in AWS East, you can use Actifio Sky running in AWS to have a local backup copy and replicate to Actifio Sky running in Azure West US to have a DR copy… or vice versa. Note that cloud vendors typically charge for data egress, i.e. any data that goes OUT of the cloud. For example, assuming you have 30 TB of production application data with a 2% daily data change rate, 50% deduplication for changed data, and $90/TB for data OUT charges, monthly costs to replicate data to a different cloud provider for DR would be approximately $810 per month. That’s the premium you could pay for the peace of mind that your mission-critical application data is safe and secure in a totally different cloud than production.
In summary, the adoption of public cloud to drive speed, agility and potential cost optimization is inevitable. Resistance to using cloud is futile. However, you have to build the right architecture, platform, and tools to ensure that your application data is safe, secure, protected, and readily accessible whenever you need it.