Through the course of our backup and disaster recovery practice, I’m asked often about the cost and complexity of modernizing a client’s backup infrastructure. My inevitable response is, ‘Remember; it’s not about backup – it’s about recovery.’
Some things to remember:
- There are two kinds of storage – that which has failed, and that which will (or, to paraphrase ‘Fight Club’: “On a long enough timeline, every disk’s lifespan drops to zero.”)
- You can back up any amount of data – but unless you’ve conducted a restoration test against both your proposed recovery time objectives (RTO) and recovery point objectives (RPO), the notion of backup by itself is meaningless.
- For a full disaster recovery strategy to be of any value, management has to have a solid idea of the costs involved. Fortunately, although complex, applying costs to both the infrastructure and the events themselves isn’t challenging.
- Data doesn’t really exist unless it’s in two places at once.
Begin at the Beginning
Reviewing your old plan and putting a rubber stamp on it isn’t enough. You may feel that ‘don’t fix it if it isn’t broken’ is a good strategy – but you may be throwing money away; insult to injury, you may be getting far less than you’re paying for.
Are you up to date on new technologies? By example – 40%+ of all SMB’s are using cloud backup in some form – but there are cost considerations which don’t readily meet the eye, and cloud backup may not be compliant with government regulations in all cases. Get a handle on both new technologies and current regulations before saying ‘yes’ to the current plan – or writing a new one.
Just as no tree grows to the sky, no budget is bottomless. Even so, there are some absolutes which ought to be considered the ‘price of admission’ for protecting data:
With 80% of businesses going under if they experience a catastrophic data loss which cannot be recovered in under four days (and that number going to 90%+ if the timeframe is a week or longer), then the RTO must be far shorter than that – ideally; 24 hours or less. It’s safe to say that no reasonable RTO is achievable without on-premise/secondary disk being the primary target for data/image backup.
On-premise disk is very inexpensive for small volume installations; conversely, it becomes prohibitively expensive for server farm images and large data backup/recovery. Unless the budget allows for this, some form of hybrid DR solution is a necessity.
With #1 and #2 being true, on premise disk is where your DR strategy should begin.
Topologies – Take Your Pick (But Remember To Test)
Some figures bear out that most IT installations are already engaged in hybrid backup/DR, even if they don’t recognize it as such. Around half of all installations still use tape; about a quarter use it exclusively. While cloud storage is the fastest growing segment of the industry, only around one fifth of SMB installations incorporate cloud as a permanent aspect of their DR topology (more on that later).
Sadly, less than 15% of all installations have ever tested their backups – and around half that number have ever simulated a complete failure-and-recovery scenario. This is a bet that no business should be taking.
The good news? Multi-platform, hybrid DR strategies are a solid approach. The bad news? A little over half of all secondary storage requirements (anything other than on-premise disk) are expected to be exceeded in the next year.
So, what are our options?
Option #1 – Physical Replication
The easiest – and probably the fastest – means of data protection is replication. It’s possible to establish a replicated environment quickly using software and existing storage, if your organization has multiple locations. It may mean buying additional storage, but the overhead is usually only an incremental expense.
(Remember, it’s almost impossible to justify this method unless deduplication is a built-in feature of either the backup software or the target hardware. A common scenario to achieve this is the use of virtualized appliances.)
Option #2 – Cloud
I often tell my clients that there are four costs to using the cloud – these costs are minimal and easily justifiable with small datasets; conversely, they become very expensive with installations much above 2TB.
The four costs are (1) the cost-per-MB to upload; (2) the cost to store; (3) the cost-per-MB to download, and (4) the cost of the infrastructure (bandwidth) you’ll use to implement #1-#3. Without a clear handle on these costs, you won’t know how much your DR strategy will cost you until you see the bill – which could be a big shock.
The other consideration is compliance – depending on the regulations governing your business (and the input from your auditors), cloud storage may not be an option at all!
Option #3 – Lower-Cost Media
Remember that almost-50% of SMB’s who still use tape? There’s a reason for that – it’s still the least expensive medium for data storage (an added benefit is that the data is still under your control). The efficiency of LTO-7 has decreased the failure-rate of tape to the point where it’s on par with disk. Coupled with a significant increase in speed, and tape is a clear winner in the secondary-storage space. (If you wondered why the major backup/DR software companies still support tape, why it’s still popular, and why many VARs, including us, still recommend it – this is the reason why.) Further, tape can be an excellent near-line solution based on preset backup and restoration policies which are built in to most backup software and appliance-based products.
Toward a Better Solution
Applying the principles above, it’s easy to see that you can create an ideal balance between expensive but fast storage (used for on-premise backup of critical business data and server images); inexpensive storage (used for near-line storage of moderately-critical data), and cloud based storage for archival of data which will rarely, if ever, be required, but which must be retained.
This is achieved using a policy based backup/DR product, whether software-only or appliance based. (Note: If some of you are thinking, “Hey! This reminds me of the Hierarchical Storage Management on a mainframe!”, you’re right. The only difference is that you can achieve the same reliability and performance at a fraction of the cost.
- There are two kinds of disk – that which has failed, and that which will. You will not be exempt from this rule.
- Disk is the most expensive form of storage. Put your critical data there.
- Use secondary storage for near-line requirements, based on preset policies.
- Cloud storage can actually be more expensive than on-premise disk, depending on what you put there. Never use it for critical data – use it for information you need to save, but will likely never need to get back.
- When upgrading your backup and DR software or appliances, make sure you’re choosing an easy to configure, policy based product which will enable you to implement a cost-balanced, yet RTO/RPO-responsive DR plan – it’ll get you where you need to be, at a price point which will make you a hero to the C-suite!
(Cymbidium Systems is a Pacific Northwest based value added reseller with a 35 year legacy of assisting small and medium business. They have practices in data protection, security, and general networking.)