Virtual Disaster Recovery Solutions

Before we further discuss disaster recovery solution, here’s a small piece of advice. Take advantage of the tools offered by leading virtualization vendors such VMware and Microsoft to help promote intra-site high availability. This should always be your primary recovery option -- cost, access to a second site and skills permitting -- as it is always faster to first failover locally whenever you are able.

Now that we have a context for a VDR solution, let’s consider some specific challenges for Virtual Disaster Recovery. Disaster recovery comes into play when the local failover capability is not an option, i.e., all of the servers are effected or the failover capability, well, fails. These unplanned events can be caused by natural disasters, such as hurricanes and floods, but are more often caused by incidents due to human error.

Without having power, or not enough reliable power (think rolling power outages prevalent in California); the best option is to utilize an available second site in an alternate geography which will not be impacted by regional power outages or adverse weather conditions.

True disaster recovery mandates that a “copy” of an organization’s information resides offsite and suitable IT recovery platforms are quickly available for use.  This “copy” can range from daily tape back-ups, periodic vaulting across the Internet, or real-time data replication across robust network connections – with these options providing (in order) decreasing RPOs and RTOs and increasing cost and complexity profiles.

As many organizations deploy virtualization to increase application agility and availability, consider real-time data replication as it best mirrors these objectives. Solutions such as VMware’s VMotion, and HA and DRS products and capabilities do a great job of managing and moving VMs and their data within a common data center location. However, they cannot by themselves replicate the data to a remote location that serves as a disaster recovery site. Therefore, a remote data replication solution must be integrated with the organization’s hypervisor (VMware ESX or Microsoft’s Hyper-V, for example) to get the data offsite. There are a wide variety of data replication solutions, with varying degrees of integration that can augment the example hyper-visor products, but most fall in to one of two categories: server/host based or storage/array based. VMware has integrated its Site Recovery Manager with leading storage/array based replication solutions.

As one might expect, each of these vary in cost/performance levels and extent of support for various operating environments. At one extreme, server-based replication solutions are the most affordable, cater to Windows and Linux environments and are best suited for small to mid-size workloads. For mid-range performance and pricing, appliance-based replication solutions are very flexible since they may have multi-vendor storage support, diverse operating systems and offer scalable performance. The appliance-based solutions generally require the client to use SAN storage, rather than internal or directly attached server storage.

Storage or array based replication are at the other extreme and offer the highest performance and price tag. Storage-based replication is usually specific to a given storage vendor’s architecture – and offers no support for heterogeneous storage environments. However, only storage replication solutions can span the complete operating environment spectrum, from small windows servers to the largest mainframes. The catch is that all these operating environments need to be serviced by the same SAN architecture.

Remote data replication solutions are reliable and well-established. Third party providers now offer the “replication engine” and second target site as hosted solutions. A hosted second site tuned for both the replication method and virtualization can take the complexity and capital expenders out while helping to meet application RPOs and RTOs

Virtual Disaster Recovery Conclusions

So how does one arrive at a DR solution for their specific virtualized machine environment?

• Find a second site to act as your DR site. Disaster recovery mandates a secure and suitably remote second site.
• Quantify your application RPOs, RTOs, and operating environment and performance requirements.
• Consider the degree of integration and support between your deployed hyper-visor and the available data replication solutions.
• Examine your VM application’s workload requirements and how powerful of a remote data replication engine is required. In addition to your VMs, if you have non-Windows or non-Linux native applications, note that server-based replication is very limited for those environments.
• If you are using SAN technology for your VM production – then you have plenty of remote replication choices.  If not, then you are limited to the server-based data replication solutions for your VMs.
• Ideally, one would like the same data replication solution to work equally well for both VMs and native applications.  Most data replications solutions do support both application environments however the extent and degree of VM support varies.
• If you are running into challenges meeting any of these, consider a service provider that can supplement skills, provide a geographically separate second, and minimize capital expenditure.

There are plenty of considerations and trade-offs to be made when selecting the best disaster recovery solution for your physical and virtual environments and disaster recovery plan objectives. The good news is there are many options that can effectively meet your objectives, as well as providers which offer pre-integrated solutions combining virtualization and remote data replication technologies to make virtual disaster recovery feasible and practical for organizations.

Related Links:

SunGard