Enterprise backing up is painful and costly: why not eliminate it?
A whole industry, along with copious processes, exists in IT to deliver backup. Much of this is far more human effort intensive (and costly) than most CIOs (or CTOs) wish to believe.
Yet backup is not the real objective: recovery is. Recovery of what has stopped working is the raison d’être for backup. Backup is pointless if the associated recovery does notfunction as intended (there is nothing new about this).
The real question is – can backup and recovery be displaced or eliminated by thinking out of the box?
Might something like XIV deliver a different sort of approach?
In 2008 IBM bought an Israeli company called XIV which specialized in making storage both simpler and more reliable for the enterprise user (or so it claimed). Many were surprised at the time that IBM made such a purchase: after all, were not storage (and backup/recovery, by implication) not already long-term IBM specialties with associated deep revenue streams?
However, when one digs deeper into what XIV provides (it is now sold as the ‘IBM XIV Storage System’), much becomes clearer. The foundation of the XIV technology is not unduly remarkable: it is a blend of:
- commodity hardware, especially standard 2TB SATA drives and associated disk and network controllers along with up to 240GB of cache
- commodity packaging, in standard modules incorporated in standard racks, with 30TB-160TB already available (going up to 240TB on 3TB drives by the end of 2011)
- software, to control the whole and to provide automated redundancy.
As you might expect, the secret sauce is in that software. This delivers the following characteristics:
- RAID-based storage
- load balancing
- redundancy at all levels (including power)
- multi-host connectivity
- automated data distribution (that self-adjusts to fit available resources/workloads)
- automated virtualized layout on disks
- automated redistribution when additional storage added
- automated recovery via disk rebuild if a component fails, including redistribution to re-establish the temporarily lost redundancy
- snapshot capabilities that do not impact performance
- mirroring/synchronizing with other XIV instances (for example, for disaster recovery)
- customer-definable QoS (if needed).
The essence of the XIV approach is that the customer needs to do (can do) little. The software is sophisticated in design, yet reliable enough to reduce manual intervention to a minimum. The evidence, from multiple customers over several years, is that XIV delivers on what is promises.
Now re-assess if you need backup/recovery
If you accept that XIV (or something like it) can deliver as described, why do you need backup and recovery any longer? If the data is automatically organized and replicated so that nothing can be lost plus it can be mirrored or synchronized at a disaster recovery site, why are all the cumbersome generations of backup/recovery plus associated manual processes required? They are not.
While this approach is not unique (though XIV’s particular implementation is), what is puzzling is that now there are manifestly reliable other ways to achieve what backup/recovery was invented to do at the enterprise level, why are not more organizations adopting approaches that eliminate so much storage and manual or semi-automated activities?
(If you have any explanations, answers or suggestions, I would be delighted to hear them.)