When it comes to protecting data, everyone really should be thinking “hybrid” first. Each medium has advantages and disadvantages correlating with different data recovery goals.
Disk, tape, and cloud: All three backup mechanisms have their strengths, and all three should be used appropriately. Iron Mountain is one IT vendor that offers a portfolio of capabilities well-matched to all three of the media types
Hi, I’m Jason Buffington, the Sr. Analyst at ESG covering Data Protection. I’ve been a “data protection dude” for 23 years, using just about every backup, snapshot and replication technology that you might imagine at one time or another. So, picking the right technology is something I’m passionate about. And I’m very bullish on hybrid backup as an architecture.
For example, think about what disk is really good for: fast backup, fast recovery. Its weakness, on the other hand, is economics – even with dedupe, disk can be expensive, compared with some other alternatives.
That reality drives people to use disk for near-term backup and recovery – i.e., they use disk for what disk is best for – but many of them still use tape for longer-term retention or preservation of data, through various backup or archiving mechanisms.
The cloud has some of the same limitations as disk, in that it’s great when you first start using it and have small volumes of data to deal with. But as your environment grows, the economics can get more difficult to justify.
Many will eventually hit a tipping point where they may be forced to reevaluate how you’re doing things in regard to your backup architecture.
Pure-play Cloud for Backup?
Not Such a Great Idea Here at ESG, we recently published the findings from our research into Data Protection as a Service. One thing that we found was that from a downtime tolerance— from an RTP/RPO perspective – almost nobody should go “pure-play” meaning straight from production server disks to the cloud route.
Here’s why – your SLAs don’t change just because you send your backups to the cloud. If you’re like most organizations these days, your SLAs are SO SHORT – your organization’s downtime tolerance is SO LOW -- that if you have to do a bulk restore across the wire, you will miss your SLAs.
My recommendation is that folks of almost any size data center should be using a local, on-prem solution (appliance or whatever) to create a disk-to-disk-to-cloud solution – from production storage, to secondary disk on-site, and THEN to the cloud repository. The onsite intermediary appliance acts as:
Well, ESG’s research shows that IT organizations, on average, need to store their data for six years. But the average BaaS provider offers data retention for one year. If BaaS vendors won’t hold your data long enough for you to comply with retention mandates, you have to keep using tape.
Of course, if you do find a BaaS provider that will store your data for the six or seven years that you require, guess what media they’ll use?
Remember, cloud-services is all about doing what you could do – at scale, and cheaper. So, how do they (or you) store data for long periods, economically? For most of us … using tape.
By the way, Tape isn’t the four-letter word that you think it is. Why is it people seem to think that while everything else has radically evolved in IT over the past 20 years, tape is somehow exactly as bad as it was in 1993?
You still need a long-term retention tier (which is likely tape for most of us), and you still need a recovery tier (which is hard to imagine being anything other than disk, preferably highly deduplicated disk).
I would offer that cloud is great for distance-based recovery and access. It a way to access data from a secondary facility as a DR or a recovery tier when disk does not work: for example, when you’ve lost access to your primary site.
And if we think not just about Backup as a Service, but also Disaster Recovery (DR) as a Service, then the cloud starts getting SUPER interesting – as a way to stand up the ability to re-start whole servers or services using a pay-as-you-need-it cloud infrastructure, instead of maintaining your own secondary site.
The overlapping zone in this diagram will look different for everyone because it reflects your organization’s unique SLA requirements and your organization’s unique cost considerations.
To be clear, I really, really like the idea of backup as a service for all kinds of scenarios – endpoints, remote offices, data centers without secondary facilities, and as an enablement for DR as a Service. But if those folks that are looking through overly optimistic lenses think that going to the cloud will relieve ALL of their hardware and software capital costs, then they could be in for a surprise. Why? Because: