Cloud DR from the Big Three: Who’s best at what?

Best for a turnkey solution: Microsoft Azure

Azure’s DR options are potentially the most polished of the three providers’ offerings.

With Azure’s Site Recovery, users can replicate physical machines and virtual machines (VMs) to a different Azure region, all from within their Azure console. It can also replicate Azure VMs from one region to another.

Azure also supports VMware VMs, Linux and, unsurprisingly, Windows physical servers. Clients can replicate these to Azure instances. Microsoft also supports replication for Azure Stack VMs, which are Azure instances that run on customers’ local hardware.

Site recovery also allows IT managers to replicate VMs on-premise or to a secondary site, using the Azure portal as well as hardware-to-cloud replication.

Azure also provides very aggressive RTO and RPO targets, which can be as little as 30 seconds running Hyper-V.

Recovery services vaults are another Azure tool. These hold data and VM configuration information which can be used to recover Azure services. The vaults work with Linux and Windows VMs, and are managed from within the main Azure console.

Not surprisingly, Azure’s DR products lend themselves to Microsoft environments, although support for other platforms is growing.

Most comprehensive DR portfolio: AWS

AWS offers several DR services, or, more accurately, services that can be tailored to disaster recovery. It also has a dedicated DR service, in Cloud Endure.

DR components include AWS Warm Standby, AWS Multi-Site and AWS Backup & Restore – although this is more accurately a tool than a service. Cloud Endure is targeted at organisations that need something close to a high availability service, but without the high costs.

There are no specific fees for using AWS technology for disaster recovery, apart from Cloud Endure.

Backup & Restore is AWS’s most basic level of disaster recovery. Users can back data up to S3, and only need to run up (and pay for) EC2 instances when they need to access data for testing or restoration.

Warm Standby works by creating a full copy of the production environment, which is always running. But it runs in reduced capacity – and so at lower cost – ready to ramp up in a DR scenario.

Organisations can also use “pilot light” mode. This is a more stripped-down operation, with just core services, that usually runs in a separate AWS region. Bringing pilot light into production will usually mean adding instances. However, it is quicker than conventional backup and recovery, and less costly than Warm Standby.

AWS Multi-Site provides active-to-active backups across separate AWS regions. This should provide zero downtime, or close to zero downtime. The penalty is the cost of maintaining a duplicated environment.

Amazon also now owns Cloud Endure. Cloud Endure DR differs from AWS Multi-Site by offering backup for physical servers and VMs as well as cloud instances. However, Cloud Endure does not run conventional hot standby systems. Rather, Amazon claims the disaster recovery technology enables businesses to launch their DR instances “in minutes” from a low-cost staging area.

This should be more cost-effective than active-to-active replication. Amazon claims low-cost staging reduces compute costs by 95%, as well as reducing the cost of OS and software licences.

CIOs should monitor storage costs, however.

Best for a modular approach: Google Cloud Platform

Unlike AWS and Azure, Google has no specific DR products. Instead, it provides detailed guidance for IT teams that want to create their own cloud-based DR environment.

Google recommends using its Deployment Manager to automate provisioning of resources, including VMs. This will bring up the Google environment if, for example, on-premise IT fails.

Google’s Compute Engine can create “instance templates” to save VM configurations. Users can then use these to launch compute instances as needed. Users can also reserve instances.

Users can then run a minimum number of instances, with minimum resources, in a similar way to AWS’s pilot light. Where businesses need to minimise downtime, they can run their servers permanently, albeit at a higher cost.

One GCP feature not currently offered by AWS or Azure is live migration for VMs, which can be useful for businesses that have a low tolerance for downtime.

Google also recommends its persistent disks for storage, as these continue to hold data even if compute instances are deleted. IT teams can use persistent disks for incremental backups or snapshots, but organisations can also opt for nearline, coldline or archival storage as part of DR plans.

Although GCP will work with the public internet, Google recommends using a dedicated interconnect to one of its colocation facilities for DR.

Much of Google’s recommended DR architecture is also available from other cloud services. And Google works with partners that package up the building blocks into ready-to-use solutions, rather than offering its own.

Google is also a strong advocate of open source technologies, so this might be attractive to IT teams that are comfortable with open source.

Picking the right Big Three DR service

Using the “Big Three” cloud providers for disaster recovery is an attractive option, because of their reputation for reliability and low cost.

However, the work needed to design the infrastructure and to replicate on-premise or existing cloud systems should not be underestimated.

At present, Azure has the most fully formed DR offering, although AWS’s Cloud Endure is an interesting alternative for enterprises looking to balance cost and the risk of downtime.

But mainstream GCP and AWS services are still very much at the DIY end of the spectrum, and that may prompt CIOs to use the cloud services’ DR-focused partners instead.