End users should consider new techniques such as "stretching" clusters, backup-to-disk with remote replication and de-duplication, remote mirroring, and virtual machines.
By Tony Asaro
Disaster recovery comes in all shapes and sizes, with some solutions costing thousands of dollars and others millions. The Enterprise Strategy Group (formerly Enterprise Storage Group) has found that most IT organizations are still using tape-based backup and recovery as their only disaster-recovery solution.
Other popular data-protection methods include the following:
- Storage array-based snapshot software is widely deployed and offers rapid recovery of data while minimizing impact on storage capacity;
- Host-based remote mirroring is extremely popular, offering inexpensive software that allows users to choose which servers they want to protect;
- There are a number of appliances and SAN switches that offer heterogeneous remote mirroring and can scale from small/medium storage networks to enterprise-class environments;
- Storage array-based remote mirroring is prevalent and offered by virtually all storage systems vendors; and
- There are also new solutions such as continuous capture and virtual tape library systems that offer new levels of data protection.
All of the above solutions have been written about to a large degree. This article discusses existing and near-future methods to protect your data that have not been overly examined in the past. These solutions may help to minimize costs and have an immediate tactical benefit toward recoverability, as well as longer-term implications on how storage networking and disaster recovery are implemented in the future.
Traditional storage systems have active-active controllers that are two physical devices but operate as one logical system. This type of architecture allows a disk array to keep operating if one controller fails. There are new storage systems that have distributed clustered architectures with more than two controllers that act as one logical system.
This allows for even higher redundancy using three- and four-node clusters to create a single logical storage system, with some configurations supporting more than 32 nodes per cluster.
There are a number of benefits of clustered architectures, including pooling resources for greater performance, ease of management, and lower cost. Another less obvious benefit is the ability to create a disaster-recovery solution by "stretching the cluster" (see Figure 1).
Figure 1: Stretching a cluster provides protection from isolated site failures such as power loss, fires, and floods.
Users can put controllers in one physical location and controllers within the same storage cluster in a different data center that is within the same campus or building (due to Fibre Channel and Ethernet distance limitations). Stretching a cluster in this manner protects users from isolated site failures, including fire, floods, and loss of power. The advantages of this solution are that it is inexpensive and relatively easy to set up.
B2D with remote replication
The popularity of backup-to-disk (B2D)—also referred to as disk-to-disk (D2D) backup—has been growing recently due to the availability of high-density, low-cost disk arrays based on ATA and Serial ATA (SATA) disk drives. There are a number of solutions that are optimized to back up data to disk offering ease of use, reliability, and performance advantages versus traditional tape-based backup/recovery.
An emerging strategy is to combine B2D and remote replication, allowing users to make copies of their backups off-site to another B2D device. This type of solution does not offer fast fail-over but does provide data protection as an alternative to archiving and vaulting tapes, which provides faster recovery times and potentially greater reliability than tape.
B2D and de-duplication
Figure 2: Using backup-to-disk (B2D) can significantly reduce the amount of capacity required for backup data.
Often, the same data gets backed up over and over again. ESG Lab has tested B2D solutions that support de-duplication of backup data, and the results were startling. We were able to consistently get a 20-to-1 reduction in the amount of storage capacity being used for backups, and we spoke with several users who were experiencing the same results. Some users were able to back up 20TB of data and only take up 1TB of storage capacity (see Figure 2 on p. 27). Your used capacity and storage costs are both significantly reduced. De-duplication also minimizes the amount of data that has to be replicated to a remote site, which also reduces bandwidth requirements. ESG believes that de-duplication will become more widely deployed in B2D solutions over the next 12 to 18 months.
Remote mirroring OS and application images
Many IT organizations store their operating system (OS) images and applications on their SAN-based storage devices. By doing this, they can use remote mirroring software to make copies of their data and their OS images and applications. This strategy makes it easy to get data centers operational after a disaster occurs and can keep costs down since a duplicate set of servers is not needed at the other end of the wire. Users will need to negotiate with their server vendors to have spare systems at the disaster-recovery site that support the same OS licenses, and this solution is not appropriate for sites that have to have immediate fail-over.
Virtual machines (VM) create system images (OS and applications) independent of the physical servers, which has powerful implications for disaster recovery. The potential is for companies to mirror virtual images of their servers to remote locations—and get them back up and running on any machine (as long as it has the same processor type).
The VM is stored on the storage system in Data Center One and is mirrored to Data Center Two (see Figure 3). There is no need to have a duplicate set of servers or applications at Data Center Two. If a disaster occurs at Data Center One, then new servers will be brought online using the VM images stored on Storage System B.
Figure 3: Virtual machines create system images (OS and applications) indepen-dent of the physical servers.
The servers can have any configuration as long as it is within the same processor family as the servers in Data Center One. This allows users to inexpensively create a disaster-recovery strategy that offers OS, application, and data recovery without worrying about the server infrastructure.
Virtual machines could change the overall computing environment and help to enable a truly liquid data center and re-invent disaster recovery in the process. This technology isn't available yet, but vendors are taking steps in this direction.
Disaster recovery is being implemented on all levels, from just using RAID and performing backups to creating identical data centers running the most sophisticated (and expensive) technologies to ensure instant fail-over and recovery. There are a myriad of technologies, products, and strategies to consider, with each providing different levels of protection, complexity, and cost. In the end, it comes down to weighing business risks versus operational budgets. The good news is that there are plenty of options; the bad news is that there are plenty of options.
Tony Asaro is a senior analyst in the Enterprise Strategy Group's ESG Lab operation (www.enterprisestrategygroup.com) in Milford, MA.