Perpetual Motion EnterpriseRock-solid storage infrastructure is the key to recovering business-critical information after a disaster
By Arun Taneja Continued from Page 1 Disk-to-disk backup and recovery. Disk-to-disk backup products can take you closer to a disaster recovery plan, depending on whether the backup is done locally or moved to a remote site. These products do continuous backups: They start with a complete backup to disk, and thereafter only transmit changes to the remote disk systems. Because these products are file-aware, recovery can be achieved at a file level. They are particularly useful in recovering files and documents that have a time association (versions). Full recovery is achievable up to the last transmitted transaction. Whether you do this backup locally or remotely, you will generally back up the secondary disk to tape.
Disk-to-disk backup will become much more prevalent, especially with the advent of iSCSI technology that allows block-level SCSI traffic to move over an IP network. Many customers will soon use iSCSI to move data from expensive, highly available, high-performance RAID systems to less expensive disk systems for backup purposes. But tape is here to stay for a long time. DISASTER RECOVERYNow we come to the final step: protecting the data center itself. There are several strategies here. When you have established which data is crucial to protect, how much new data is generated each day, how quickly you need to recover after a disaster, how much data you are willing to lose, and how much you are willing to pay, the right solution will present itself. Synchronous mirroring or replication. Two methodologies are possible — synchronous mirroring (or replication) between two or more storage systems, and replication via servers. EMC, IBM, and Hitachi Data Systems offer software products in this category that support the replication of all changes to a volume in the primary storage system to a volume in the remote storage system over Entreprise Systems Connection (ESCON), Asynchronous Transfer Mode (ATM), or IP networks. The servers (hosts) are not involved in the transfer. There is no files ystem awareness involved; changes are made at the block level and recovery occurs at a volume level. Changes are not written to the primary volume unless they are written on the remote volume and an acknowledgment received to that effect — an approach that affects the performance of applications running on the primary site, worsening with distance. The performance is also proportional to the speed of the network deployed between the storage systems, but data integrity is preserved. Transactions are consistent on both sides; if the primary data center is lost, the applications can run on the secondary site using the mirrored volume. Except for the performance impact, high cost, and potentially long time-to-recovery, synchronous mirroring is the safest solution for disaster recovery. (Note that you need like storage on both sides.) The solution can also be designed for one-to-many replication, but this strategy is expensive. Manual switching to the remote servers is required unless wide area network clustering is employed. Synchronous mirroring can occur in the host as well. In this case, the changes to a volume are sent via the server over the network to the remote server and then to the secondary volume. As before, the transaction is not completed on the primary side unless the remote side writes the changes and acknowledges receipt. Storage and applications performance are affected, and longer distances exacerbate that impact. However, one advantage of this approach is that you can use different storage systems on the primary and remote sites. For example, you could use a highly available, high-performance disk system on the primary side, but a less expensive system at the remote site. In any case, the remote site needs more than just the storage system: It needs a nearly identical server and network environment so that applications can be started immediately after the disaster. Because the data flows through the filesystems, recovery can be achieved both at the file or a full volume level. Another advantage is that you can invoke one-to-many replication easily and relatively affordably. Asynchronous mirroring. If the performance penalty for synchronous mirroring is unacceptable, you can opt for asynchronous mirroring. In this approach, changes are recorded on the primary site as they occur and simultaneously "thrown" on the wire. They travel over the network and are recorded asynchronously on the remote volume. Although the performance impact is minimal, the remote site can be out of synch by the amount of data in the funnel, so the longer the distance, the worse the consistency factor after a disaster. You may find that while this approach is unacceptable for the business-critical applications, it may be totally acceptable for others. Disaster recovery used to be the domain of only the largest enterprises, but with asynchronous mirroring, even smaller companies can start to seriously consider disaster recovery plans. These solutions are not expensive and having access to all but a few seconds of data is much better than the alternative. RECOVER AND PROSPERWe hope and pray that events similar to Sept. 11th will never reoccur. But Mother Nature has a way of unleashing her fury in the form of floods, earthquakes, and other calamities, and human error is common to all businesses. A storage infrastructure built on a solid foundation will go a long way toward keeping your business humming in spite of these ravages. Arun Taneja [arunt@enterprisestoragegroup.com] is a senior analyst with the Enterprise Storage Group, an analyst group focused on storage technologies. He has 25 years of experience in the storage and server industry.
|
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
| |||||||||||||||||||||||||||||||
