By Dave Bermingham
No infrastructure is wholly impervious to disaster. Wildfires, floods, even simple human error can bring down even the most well-designed data center. So how can you minimize the risk of operational disruption and data loss as a result of a disaster if, for whatever reason, you choose to rely on an on-premises infrastructure?
Many organizations facing this question are opting for disaster recovery (DR) solutions built on failover clustering technologies. In a separate, geographically distant location, they assemble the resources they require to run copies of their critical applications and databases at production scale. They rely on a data replication solution to ensure that their critical production databases are copied to that distant DR site – in a manner as close to real time as possible – so that they can spin up that DR site and be operational quickly if the primary on-prem infrastructure were to go offline.
But physically replicating your on-prem infrastructure in a location halfway across the state or country can be a costly solution. The investment in a proper data center, in hardware, software, and support resources may be prohibitive. Moreover, while the DR site you build today may reflect the infrastructure you need today, it may not reflect the infrastructure you’ll need next year; and regularly upgrading that DR site to ensure that it’s ready to be called into action – for a disaster that may never arise – is going to be even more costly.
An alternative approach to DR involves building a hybrid failover cluster, one that connects your on-prem infrastructure to a cloud-based DR infrastructure. This approach can provide the same level of operational continuity that you could secure using a failover cluster spanning two separate physical infrastructures, but it can also do so in a way that is less costly, more flexible, and more scalable.
Disaster recovery and high availability: similar but different
High availability (HA) and disaster recovery (DR) are two essential strategies for ensuring business continuity, but they address different challenges and operate differently. While both aim to minimize downtime and protect data, the key distinctions lie in their configuration, purpose, and operational mechanisms.
High availability
HA solutions are designed to ensure continuous access to applications and data within a specific environment, even in the event of localized failures. In a HA configuration, the primary and secondary infrastructures are typically located near one another. This geographic proximity enables the use of synchronous data replication mechanisms, whether application- or third-party-based, ensuring that every transaction written to the primary storage is simultaneously written to the secondary storage.
If the primary infrastructure fails, failover to the secondary infrastructure is automatic and seamless, with no loss of data. HA configurations are ideal for mission-critical applications whose recovery time objectives (RTOs) and recovery point objectives (RPOs) demand immediate failover and no data loss.
Disaster recovery
DR solutions, on the other hand, focus on recovering from large-scale disasters that impact entire regions, such as wildfires, storms, or earthquakes. As such, the primary and DR infrastructures should be located in geographically distant regions to ensure the secondary site is unaffected by the same disaster.
Due to this distance, data replication in a DR configuration is asynchronous. Transactions written to the primary storage are replicated to the DR storage as quickly as possible, but the two systems may not always be perfectly in sync. This introduces the potential for a small data gap (several seconds or even minutes) if the DR site needs to be brought online. Despite this limitation, DR configurations provide faster recovery and less data loss compared to restoring from an older backup.
Failover in a DR solution is typically a manual process. While it’s possible to configure automated failover in cases where all local HA nodes are lost, many organizations prefer a push-button disaster recovery approach to ensure proper evaluation and control before activating the secondary infrastructure. This manual activation helps prevent accidental failovers and ensures that the business is prepared to resume operations from the DR site.
By addressing different levels of risk, HA and DR complement each other to provide comprehensive protection for business-critical systems. While HA prioritizes seamless availability and data integrity within a localized environment, DR focuses on minimizing downtime and data loss in the face of catastrophic events.
The benefits of a hybrid approach to DR
So why put your DR infrastructure in the cloud? To begin with, you don’t need to invest the time or acquire the human resources to build out the DR infrastructure. You can work with Google Cloud, Azure, AWS, or other specialists to configure your cloud-based DR solution, and you can provision all the IT resources you require – from load balancers to memory, cores, and storage space – in a matter of minutes. You can link the on-prem and cloud-based infrastructures using the same native or third-party failover clustering and data replication tools that you would use if you were using a physical DR site.
Moreover, if you need to scale up your DR infrastructure because your actual on-prem infrastructure has evolved – or because you intentionally under-provisioned your DR site to reduce costs while it was in standby mode – you can scale up your DR solution instantly with a few clicks in a browser window. That way, you can ensure that your DR site is always ready to provide the right level of support in the event of an emergency.
Considerations when configuring a hybrid DR solution
One key consideration when configuring a cloud-based DR solution is your network. The tools you choose to replicate data among the nodes of your failover cluster will be moving a lot of data to the cloud, so you’ll want to configure a secure high speed virtual private network (VPN) or dedicated interconnect to your cloud-based infrastructure. This will minimize latency and maximize data throughput to ensure that the data in the cloud is as in sync with the data in your on-prem databases as possible.
Similarly, be sure that the data replication solution you select is replicating all the data you need replicated to your DR site. Some application-native replication solutions replicate only the data associated with that application. Replication systems untethered to specific applications can replicate all the data in a specified storage location, regardless of the application that wrote the data to storage. You don’t want to find that data you thought was being replicated is not present on your DR system when it matters.
The author
Dave Bermingham is the Senior Technical Evangelist at SIOS Technology. He is recognized within the technology community as a high availability expert and has been honored by his peers by being elected to be a Microsoft MVP in Clustering six times and seven times as a Cloud and Datacenter MVP. Dave is a frequent speaker at technical conferences, including SQL Saturdays, Pass Summit, and MSSQL Tips, and is the author of Clustering for Mere Mortals blog. Dave holds numerous technical certifications and has more than thirty years of IT experience, including in finance, healthcare, and education.
