Disaster Recovery in the Cloud: How New Backup Technology is Changing DR Planning
Tape backup is still an important component to a good backup strategy, but new technology means we need recovery options that are faster than tape can deliver. Here's why you should consider a mix of solutions. Last in a series.
- By Eric Beehler
In order to execute a disaster recovery, you have to rely heavily on your backup solution. Not only do you need to be able to grab accidentally deleted files or recover a crashed server in your day to day duties, but you also have to restore your entire infrastructure and critical applications. This puts a new requirement on your backup solution that doesn't exist in the normal scope of typical backup responsibilities -- a need to get multitudes of systems up and running quickly and without data loss.
The tried and true tape backup and restore solution is becoming passé. Today's business requirements don't really provide for an outage of hours, or even days to restore critical functionality that most tape solutions require. With the newest backup technologies offering whole machine backup and restore and instant replication of databases, you have much better Recovery Times and Recovery Points. These new technologies are great but much more resource intensive, so with that has come more complex planning to successfully restore and recover.
Offsite Disk-Based Backup
Many companies are now using disk-based backup to reduce the backup window and pushing tape out to off-site storage. This is a natural progression, but too many companies are relying on this as their disaster recovery solution without proper planning. With the backup tapes usually containing at least two- to three-day-old data, the need to spool that data back to servers, and the continued challenge of building up servers and networks to re-light those applications, there is still plenty of room for improvement. There is still room for tape, but only for your applications and servers not deemed Tier 1 at the highest levels of RTO and RPO.
When using disk-based backup, there is usually an option for a replication solution. This is usually a continuous data protection scheme, which means you have a point-in-time backup that is replicated to the recovery disk at another location. This is similar to your tape backups, but without the physical media handling and seek-time needs of tape.
With a disk-based backup, deduplication is integrated into many solutions, to remove duplicate files and replace them with pointers, which can increase disk space utilization dramatically. For disaster recovery, it's essential to push smaller amounts of data over a WAN connection, but you must consider the time it takes to "rehydrate" the backup from the deduplicated state, which will factor into your time to restore.
Could Mirroring Be the Solution?
Some solutions are so real-time that they fall outside the abilities of backup software. Critical databases should be replicating transactions in real-time. Storage vendors have integrated features to allow for block-level copies at real-time as changes are committed, which is often labeled as some variation of mirroring. This allows you to avoid the file-level focus of some backup software and only copy the bits that have changed, increasing efficiency for a WAN-based mirroring solution even more. To effectively utilize this kind of replication many vendors are integrating WAN deduplication to reduce the amount of data stored and transferred by a multitude of 20x or more.
Virtualization is providing the hottest trend in disaster recovery by utilizing the ability to take a snapshot of a virtual machine and have it running on a different server at a different location. Virtualization-specific disaster recovery allows you to have your production network running at an alternate location, reducing the old pains of reconfiguring new servers to run software already working in your primary data center. This is a great solution, but must be balanced with the other pieces of your infrastructure. Storage-, network- and application-specific dependencies are not necessarily covered just because your virtual servers can be back online, but introducing additional automation to hasten the march towards a faster recovery is a great addition to your DR portfolio.
As opposed to what's being sold as neat all-in-one packages, you'll likely need to use a combination of mirroring for storage, continuous data replication for flexible restore points, WAN optimization, deduplication in disk-based backups, virtualization-specific replication, and good old tape recovery to help reduce costs for less critical systems. This is why DR backup has become complicated, but at the same time is much more freeing since you have so many solutions available.
Just remember that what works for a virtual machine may not work for your databases, and what works for your databases may not work for your big file servers. The list goes on, so rely on your understanding of your services and their dependencies to set up the proper backup solutions for each system.
Eric Beehler currently has certifications from CompTIA (A+, N+, Server+) and Microsoft (MCITP: Enterprise Support Technician and Consumer Support Technician, MCTS: Windows Vista Configuration, MCDBA SQL Server 2000, MCSE+I Windows NT 4.0, MCSE Windows 2000 and MCSE Windows 2003). He has authored books and white papers, and co-hosts CS Techcast, a podcast aimed at IT professionals. He now provides consulting, managed services and training through his co-ownership in Consortio Services LLC.