Cloud PRA: which hyperscaler to choose

Public cloud hosts known as “hyperscalers”-Amazon Web Services (AWS), Google Cloud Platform (GCP) and Microsoft Azure-are now big players in data storage, archiving and backup. They also offer Disaster Recovery Plan (DRP) options.

Before the advent of the cloud, a DRP involved either purchasing and putting into production redundant infrastructure to install them in a second data center, or using dedicated services from a subcontractor.

The public cloud offers the potential for a flexible and cheaper service. Disaster recovery specialists respond to large businesses in industries with low error tolerance. But a company can create its own example of PRA in the cloud using just a web browser and a credit card.

And because most CIOs prefer at least dedicated DRP support and services than an in-house DIY approach, public cloud providers have adapted their IaaS offerings to deliver turnkey DRP solutions. .

Therefore, IT teams will need to understand the offers against the constraints of their own infrastructure before selecting a public cloud DRP solution.

Each vendor has a different approach. Therefore, IT teams will need to understand the offers regarding the constraints of their own infrastructure, before choosing a DRP solution in the public cloud. Each platform has its pros and cons.

The best for a turnkey solution: Microsoft Azure

Azure’s disaster recovery options are probably the most mature of the offers of the three major public cloud providers.

On offer Site Recovery In Azure, users can copy physical machines and virtual machines (VMs) to the Azure region of their choice, simply from their Azure console. They can also copy Azure VMs to regions.

Azure replicates VMware, Linux virtual machines and, not surprisingly, Windows physical servers. Customers can convert them to Azure instances to be returned directly to the cloud. Microsoft clearly supports replication of Azure Stack VMs, as they are Azure instances running on local hardware rented from Microsoft customers.

Site Recovery also allows VMs to be copied to the original site or to a second site. Finally, Azure offers the most interesting recovery times (RTO) and recent data restoration (RPO): with the Hyper-V hypervisor, data that dates back to 30 seconds before being restored incident within 30 seconds.

Ang Recovery Services (vaults for services to be restored, in French)-another Azure tool-keep information about the configuration of virtual machines. They are used to restore the contents of the original VMs to Azure services. Vault runs Linux or Windows VMs, always controlled from the main Azure console.

The most complete portfolio of functions: AWS

AWS offers several DRP services or, more accurately, services that can be tailored for disaster recovery. Officially, there is only one service stamped PRA: Cloud Endure.

Components that can be used for a PRA include AWS Warm Standby, AWS Multi-Site, and AWS Backup & Restore. Cloud Endure is for companies that need a service close to high availability, but without high costs.

AWS Backup & Restore is the most basic level of disaster recovery. Users determine what data they back up to S3 and read it back from this repository to restore or test.

AWS Warm Standby, it creates a twin production environment, ready to take over in the event of an incident. However, its performance is limited. We are more talking about a degraded mode.

Companies can also use the mode Pilot lights, more stripped, where only basic services can restart immediately. They run in a separate AWS Region. Switching to a Pilot Light environment is usually not enough to restart activity: companies must assign new opportunities to it so that it can be completely replaced. However, Pilot Light can restart production faster than Backup & Restore and cheaper than Warm Standby.

AWS Multisite creates a clone of the fully functional production environment, in active-active mode, with each environment located in a separate region. This option should avoid, or virtually avoid, any downtime. The other side of the coin is that the company must pay for two environments.

Cloud Endure is a recent offering. This service works like AWS Multi-Site, with the difference that it also synchronizes the content of physical servers and VMs running from the company’s datacenter. However, Cloud Endure is not synonymous with AWS Multi-Site. AWS argues that it can, however, restart activity “in minutes” and, from a lower-cost AWS location.

According to AWS, in this area, computing capacities will be 95% cheaper. Operating system and software licenses will also be cheaper. Only used storage capacity will have the same price.

The best modular approach: Google Cloud Platform

Unlike AWS and Azure, the Google Cloud Platform (GCP) does not have a dedicated DRP service. Instead, it provides detailed guidance for IT teams looking for their own cloud-based PRA environment.

Google recommends using it Deployment Manager to automate the provision of resources, including VMs. The Google environment will be active if, for example, on-site IT fails. The IaaS Compute Engine service saves configurations of virtual machines required for disaster recovery in “Instance Templates”. Then, users use these templates to deploy the required compute instances. The advantage is that companies can restart a minimum of times, just enough to work from the cloud in an acceptable degraded mode, such as AWS Warm Standby. Businesses that need to reduce downtime will choose to run standby servers at all times, but at a higher cost.

Google also recommends using its persistent disks for storage, as they still contain their data even if associated compute instances are lost. IT teams will use persistent disks to store incremental backups or snapshots.

One feature of GCP that is not currently offered by AWS or Azure is rolling VM migration, which is useful for businesses with low tolerance for downtime. Although GCP is accessible from the Internet, Google recommends using Direct Connect links in its instances located in the colocation data centers closest to the company in this case.

In addition, much of the PRA architecture recommended by Google can be replicated elsewhere, in another public or private cloud, or even in companies ’datacenters. Google says it works with partners (e.g., OVHcloud in France) that develop its core technologies into ready -to -use solutions. These technologies are Open source bricks, so the IT team needs to be comfortable with Open source to build a DRP that copies a site’s physical servers to a GCP partner’s IaaS.

An attractive option, but beware of DIY

Utilizing the big three public cloud providers is ultimately an attractive option because of their reputation for reliability and low cost. However, the work involved in designing infrastructure and copying systems on-premises or in the cloud should not be underestimated.

Azure currently has the most comprehensive DRP offer, while AWS Cloud Endure is an attractive alternative for businesses looking to balance the value of DRP with the risk posed by downtime. ‘stop.

For others, the standard PRA services in GCP and AWS are still very close to build-it-yourself solutions. They encourage CIOs to go through partners who are experts on these issues instead.

Leave a Comment