I have written extensively in this space on the current status and the future directions of the Open Virtualization Format (OVF) standard from the Distributed Management Task Force (DMTF). This standard recently achieved ANSI standard approval as well. The ability to define workloads, inclusive of virtual machine instance counts and network topology, in an industry-standard format is a powerful tool.
One area I have not explored in much detail is the ability to leverage OVF in disaster preparedness and recovery processes. Disaster preparedness can take many forms and involve many elements. Some elements are not even technical in nature, such as plans to keep executives in the decision-making process and making alternative call centers available for extended periods in extreme cases. As it concerns the data center, data mirroring and system backups on network storage is a basic solution targeting hardware attrition. Using an offsite storage solution is usually a requirement. The complexity and cost grows exponentially when we consider requirements for near real-time data center failover. Synchronizing data and system images across sites is a complex problem, never mind bringing critical systems online in a different data center. Keeping multiple data centers running is an option only for large enterprises.
One promise of “cloud computing” is making the data center backup and migration problems more tractable. Compute, storage and network resources can be abstracted from virtual machines. Workloads thus become much more portable. Capacity for all three resources can be rented on a temporary basis, potentially driving the costs of disaster preparedness down even in the most complex cases.
However, some hurdles remain. For example, how does an IT department keep the running images synchronized with the off-site backups used to recreate the cloud? For the real time failover, how does it manage the resulting data transfer requirements? How should the cloud administrators describe all the discrete components running on a cloud so that those components can be realized in a temporary or replacement environment?
Using OVF packages to describe your virtualized data center workloads creates a baseline for disaster recovery on a small or large scale. OVF includes metadata about CPU, memory, and network requirements for each virtual machine. The basic description needed to realize each virtual machine in a new cloud is therefore already included. If network topologies and instance counts are included in a single OVF backup for a group of related virtual systems, then entire appliances can be directly realized in a new cloud environment. The compression and pending inclusion of encryption in the 2.0 standard will optimize images for off-premise storage and/or transfer over WAN environments.
Inclusion of these OVF appliances in the normal process of data center backup makes restoration of one or more appliances much less costly. Even restoring an on-premise data center cloud from a total loss would be substantially easier. Admittedly, the real time failover case is more a future direction than a current reality. The relative immaturity of image synchronization and the challenges of large scale back up to a cloud staging area make implementing such a strategy very challenging and expensive. However, making such options available is a matter of time and technical maturation than anything else. Cloud providers such as Rackspace are already working to provide public cloud OVF support that could be eventually leveraged for temporary data center operations.
So, what does this mean for a data center today? Vendors such as Citrix and VMware already include OVF support in their base products. As stated, building your cloud workload definitions around OVF today already makes it easier to back up and restore those workloads on your existing virtualization environment. Perhaps more importantly, such a strategy will also prepare your IT infrastructure to reap the future benefits of cloud computing, including the option for offsite recovery.
In short, disaster recovery is yet another example that proves the relevance of OVF going forward.