By Rick Litvin
Snapshots are one of the key critical components of Enterprise in delivering data protection services to ensure adequate business continuity (BC) and disaster recovery (DR) strategies. Snapshots ensure more aggressive RPOs and RTOs values when compared to traditional streaming backups, eliminate the need to plan for and schedule backup windows, and are great fit for virtualized environments.
Despite all these great advantages of snapshots’ technology, one should keep in mind the following considerations for VM & Storage-based snapshots from both technical and operational point of view:
– Should be used for short term maintenance activities (e.g., upgrading or patching apps & servers), with minimal data change rate
– Allow quick and easy point in time rollback in case of application level failure or testing
– Manual intervention might be required by various team members, i.e., might require an additional coordination with data/apps owners for data conflict validation
– VMs & Host I/O performance impacted while running
– VM will crash if snapshots fill-up datastore and customer can suffer some corruption due to a maxed out datastore caused by snapshots. This happens when the admin has no awareness of the importance of monitoring for snapshots.
– Virtual disks cannot be expanded while snapshots are active
– Performance impact based on snapshot implementation approach
Traditional ‘copy on write’ (COW) implementations result in an instant snapshot. However, during the original data changes, the original data is copied to a reserve area, and as a result, corresponding metadata is updated accordingly. Over time, as more snapshots are taken, the performance gets increasingly worse due to this activity.
On the other hand, another snapshot implementation approach [‘redirect on write’ (ROW)], redirect writes to original data to free locations on disk. Since the original data is not copied, ROW implementation does not have similar performance impact compare to ‘COW’ approach.
– VM Consistency should be enforced by ensuring / verifying that snapshots are set up correctly to be VM consistent
– Application consistency should be considered based on requirements & strategy
– Snapshot Replication as part of Data protection approach for DR Per SNIA [Storage Networking Industry Association] … a snapshot is not really a backup until it has been replicated to another storage system.
– Storage array capacity planning process should take into consideration storage based snapshot-related activities to properly size required storage capacity based on how many snapshots should be taken in a given period, how long these snapshots should be retained (typical values are 30, 60, or 90 days retention), DR (replication) strategy, etc.
– Custom, non-standard process of manual (non automated) snapshot management will be in place if proper / adequate snapshot management technology won’t be deployed.
As indicated above, much of the effort that has been put into development of snapshot-based data protection strategy won’t be as efficient or in vain if proper / adequate snapshot management technology won’t be deployed. In other words, If companies don’t implement proper snapshot management technology, they will end up with inefficient, non-standard process.
Today, evolution of vendors’ approaches in snapshot management technologies is coming from 3 different vendors’ types:
– Traditional Streaming Backup Vendors — the only vendor’s type who developed & offer unified data protection approachwith both snapshot management functionality and corresponding streaming backup products (Example of Snapshot Management Products: Symantec ‘NetBackup Replication Director’; HP ‘Data Protector Zero downtime backup, Instant recovery and vStorage Image backup method’; CommVault Simpana INTELLISNAP technology; Veeam Availability Suite, EMC NetWorker Snapshot Management, Asigra Snapshot Manager etc.).
– Storage Vendors [E.g., EMC Replication Manager, NetApp, Hitachi, HP etc.]
Traditional storage array based built-in snapshot management does not have an ability to track and index the various snapshots. In most cases, these types of vendors can only offer reduced snapshot management functionality, limited to the technology developed by specific vendor.
– New Vendors’ Category (vendors who are not part of traditional streaming backup or storage’ vendors) [E.g., Catalogic, Actifio etc.]. These vendors do offer very wide spectrum of various snapshot / copy management functionalities, but do not offer traditional streaming backup approach.
The most advanced snapshot management technologies do offer storage array and Hypervisor integrated snapshots with the following functionalities:
– Integrated snapshots, mirrors and vaults – all within one centralized GUI
– Optimization of regular copy data use for automated test / development builds, DR testing, Analytics, DevOps etc.
– Scheduling storage-based snapshots across multiple storage arrays, applications and VM virtual servers
– Ability to catalogue snapshots for full volume or granular file recovery
– Quick search and recovery of individual files across collection of indexed snapshots
– Quickly create and retain application aware snapshots
– Actionable catalog with extensive reporting and file analytics
– Self-Service Copy Provisioning and DR Capabilities
Despite the great appeal of snapshots (clones, replicas, etc.) and corresponding snapshots management functionalities, many enterprises are still reluctant to completely design / architect their data protection strategy based on snapshot based backups only. However, there is definitely a trend to implement 100% snapshot based (clones, replica, etc.) data protection without combining with traditional streaming backup approach. According to Gartner study, by 2016 at least 10% of large enterprises will have given up on traditional backup / recovery software and will employ snapshot and replication techniques instead.
Before implementing Snapshots management / automation technology within an environment, one should very carefully consider vendor’s technology from interoperability / supportability point of view for current and future infrastructure components, functionality (various vendors, including ones listed above, do have different level of maturity & functionality of the technology), and operational readiness prospective.