Few organisations would dispute the importance of backup. For some it's a necessity driven by compliance from customers or regulators. For others it’s someone’s learned experience on what happens when you don’t back up your data. For others it’s the fear of being blamed if something happens – and Murphy’s Law says it will.
But not all backup implementations work well. Years ago, the CIO of an Asian multinational conglomerate discovered that data backed to tapes were unreliable. This happened because a disaster occurred and IT could not recover data from tapes.
So we asked several backup vendors about the horror stories they have heard over the years and to pick the worst backup strategies from among the lot. In this first of a two-part series we highlight common backup practices that should be corrected.
Number one on the list is not having a backup at all. PK Gupta, Senior Director and Chief Architect, EMC Data Protection and Availability Division, Asia Pacific and Japan (I know it's a mouthful) says this is the worst possible thing to do (or not do). He notes that large enterprises typically have a disk-based or tape-based backup strategy. But in many cases he laments that these are siloed and therefore ineffective. SMBs fare no better adding USB disk-based backup solutions which pose a “huge risk of data loss”.
Got Backup But Don’t Test
This is the sin of the Asian conglomerate I mentioned above. Yes they do regular backups but no consistent testing was done. Arshad Munir Sharif, General Manager for System & Technology Group, IBM Malaysia says most organisations have a habit of backing up their data daily but don’t perform testing of the backed data.
“This is flawed because when they experience data loss or they realise that data cartridges where the data is supposed to be stored are empty. Why? Because files are missing or the data is corrupted. The hardware being disk (VTL), tapes and drives can fail. Organisations have to perform regular recovery tests for files, databases, operating system or full system recovery to pre-empt unwelcome surprises sooner than later,” he advised.
No Consideration for Recovery
As in the case of the Asian conglomerate, it appeared that no afterthought was given on the inevitability of a restore. So at some point when a disaster did occur, the discovery and recovery process took longer than expected.
Raymond Goh, Technical Specialist, Asia for Veeam Software, adds further that having in place recovery verifications and procedures that are not well defined. This could be because the administrator has no confidence in the proper way to recover data or perhaps, in the first place, does not even know whether data can be recovered.
VTLs are not the answer
There was a time when virtual tape libraries (VTLs) were seen as the adjunct if not replacement for tape libraries. VTLs use software to mask what are essentially disks and make these systems as though they were tapes to backup software. The benefits of such virtualization include storage consolidation and faster data restore processes. But the advent of purpose-built backup appliances (PBBA) may be signalling what many have long declared as the end of life for VTLs as well as tape siloes.
Mark Bentkower, Director of Enterprise Solutions Asia Pacific, CommVault Systems is one such advocate. He says VTLs are part of an outdated strategy that position VTL as high performance backup solutions – at least when compared to physical tapes libraries.
He reckons that VTLs will continue to have a position in the datacentre as long as companies maintain old midrange systems that have fibre channel backup paths. He argues that the real beneficiaries of VTL are storage vendors.
“Retiring storage tape tiers in lieu of putting everything on VTL rarely, if ever, reduces TCO. Marketing de-duplication ratios are rarely achievable in real day-to-day business practice across the wide variety of data types that an enterprise has to manage. This ultimately leads to untimely additional purchases of VTL storage and “TCO shock” when the solution vastly under-delivers promised savings,” explains Bentkower.
Ignoring Backup Logs
Because backup is seen as a necessary evil, its day-to-day execution is often relegated to last place when it comes to importance. As long as there is no major system crashes, backups are assigned to the automated backup software and left to its devices.
Sharif said logs are important because they record skipped files, failed backups or errors in tapes or drives. He says that if an organisation fails to monitor and take prompt action this can result in failure to recover.
“For example, a database backup did not complete on time during scheduled backup because of too many changes in the data but the system has to come back to production spells danger. The backup team may have realized that the backup was not complete but did not take any action to remedy it following the failure. This exposes organisations to further risks,” he cautions.
The Right Number of Backup Copies
Is having one copy of your data sufficient? Hardly! But of course if depends on how frequent the data is being updated, the volume of updates, and the nature of the data and the applications that access it. Whilst a single copy is inviting disaster, too many duplicates can potentially cause wastage. The backup process itself takes time and whilst backup is happening, systems and applications are usually slower to respond.
Gupta adds that “many organisations still don't have automated backups and are keeping multiple copies of data. As a result, the manpower, tape management, power, cooling, space requirements continue to increase putting pressure on the IT budgets.”
Snapshots do not replace Backups
Veeam’s Goh says that SMBs typically mistake snapshots as replacing backups. It is a mistake to “utilise virtual snapshots as a form of backup without realizing that this is an obvious single point of failure on the production storage. This creates a situation where the production data and snapshots are stored in a single basket.”
There are many other commonly other practices that continue to be observed that do not reflect the modern datacentre environment.
In this second part of this series, we will look at ways to recover from a bad backup strategy as well as determining if your organisation has, indeed, a bad backup practice in place.