What you will learn from this tip: Most agree that tape backups are no longer suitable for all data protection needs. Organizations are now presented with numerous backup options to choose from. This tip will help you rationalize some of these options.
Many organizations that have suffered an outage or failure requiring a major data restore realized that their tape backup product would never cut it in the event of a site-wide disaster. Recovery time objectives (RTO) have consistently become shorter, driven by tighter service level agreements (SLA), higher end user expectations and increasing IT dependency.
IT organizations now have numerous options to choose from; however, questions come up on a regular basis regarding the differences between the many different technologies. What follows is an attempt to group and rationalize some of these options. That said, it should not be considered an extensive dissertation of all available technologies, brands and flavors.
Backup to disk
Backup to disk is readily available with many "traditional" backup products. It is essentially just another media to which the backup software can write data and includes virtual tape libraries( VTL). Disk backups offer faster performance than tape through reduced data-seek time and by eliminating media mount wait. RAID technology for disk also significantly reduces the risk of data loss resulting from tape failure. Note that in the context of this discussion, data replication or mirroring, which is often considered a form of disk backup, is addressed separately.
Data replication is a fairly broad term; functionality and capabilities will vary from one vendor to another. Some of the main flavors include:
- Traditionally more expensive than software-based products and often hardware dependent, although many improvements were made in that area.
- Offloads processing overhead from the hosts.
- Can replicate between local or remote arrays.
- Replication granularity can be at the block or volume level.
- Disk mirroring is probably the best known form of replication -- although not necessarily the best one.
Snapshot vs. volume replication
- Entire volumes can be replicated and accessed independently; can significantly increase storage requirements as each copy is the same size as the original.
- Snapshots represent point-in-time images of a volume or data set and can be made of changes only in relation to the initial copy. It uses considerably less storage than entire volume copies but depend on common data (unchanged) to offer multiple images.
- Host-based replication is software driven and typically hardware independent.
- Processing overhead is at the host level.
- Offers a lower cost alternative to hardware-based replication and supports dissimilar hardware.
- Replication can be local or remote.
Host based with failover capabilities
- Same as host-based replication but with added failover functionality.
- Replication can be part of the high availability product or used to complement existing remote failover software.
- Some vendors allow remote replication and failover to dissimilar hardware.
Synchronous vs. asynchronous
- At a high level, synchronous replication means that data is replicated to the remote site -- or onsite replica -- instantly. Both original and replica are always synchronized. When bandwidth is limited or latency is introduced by significant distance between copies, asynchronous replication is used. This means that changes to the original copy of the data are buffered or queued to be written to the remote, disk with some delay.
- Note that synchronism of replicated data is usually sought mostly for "mirror copies" or any type of replication where both copies must be constantly and consistently identical.
- A variation on the array-based replication mentioned earlier but with specific types of storage arrays.
- This is the case for many network attached storage (NAS) products that usually have their own type of operating system or intelligence beyond that of a simple disk array.
Since cost for the data protection options listed above will vary widely based on technology, vendor, platform and functionality, the next logical step is to categorize data. The right solution for each data category will be dictated by the value of the data to the organization, potential losses and recovery requirements (i.e., RTO and request for propofal [RPO]).
Do you know…
About the author: Pierre Dorion is a certified business continuity professional for Mainland Information Systems Inc.