This article is more than 1 year old

Expert view: What is the forecast for cloud backup?

The outlook is hazy

Hybrid cloud

So cloud only and cloud first have been mostly shot down. It seems like the time has not yet arrived when it is always the best option, or even most likely to be the best option. Things are looking grim in Mudville…

Is there any hope for cloudy backups? Does it at least make sense to take a hybrid approach, with some backups handled by cloud services and some dealt with in-house?

vExpert and blogger Rick Vanover seems to subscribe to Mellor's tan blob vision of cloud storage.

"Cloud storage, cloud-based backups and backups that are not administered by the end-user could be classified as a tier,” he says.

Vanover believes that much as we tier storage in data centres today, cloud storage could serve as an archive tier.

Fowler offers an important thought about hybrid cloud storage: there are cases where it makes good sense to back up from the cloud to your local network.

"It depends on the data. If you have a website or certain content that is hosted in the cloud, it can make sense to also have the cloud as a primary backup location, with secondary less regular backups done privately,” he says.

Food for thought. A hybrid approach to cloud computing is not simply a means to the long-term end of getting all your data into someone else's public cloud. A truly hybrid approach would involve deciding when paying a subscription fee is rational and when owning the kit yourself is the better call.

Sometimes you have data in the cloud, sometimes you have it on your network. Where the backups go depends on the workload in question, your risk tolerance and your existing infrastructure investments.

Who's on cloud?

What is happening in the field? Do our experts see much adoption of cloud backups?

Wright thinks more people are investigating cloud backups. “The initial driver in many cases is protecting Robo [remote office/branch office] and mobile workers,” he says.

While he believes cloud backups should be considered for larger data centres, he thinks they have a fight ahead of them as they must compete against existing hardware (effectively paid for) and a long-standing operational model.

“The incumbent solution is difficult to unseat,” he points out. Quite. Other contributors echo this tyre-kicking view.

So just who is using cloud backups? What are they using it for? Clearly there is some adoption but there doesn't seem to be a lot of top-down push towards it in many organisations.

If there is to be a sea change in attitudes towards cloud backups, it might help if we understood where it is being adopted and why.

Wright says that in his organisation there is very limited adoption of public cloud resources for backups.

“Dropbox and similar tools are in use, but not necessarily with the blessing of IT,” he says.

I am sure you can all guess why; in Wright's world, Dropbox et al are "usually used for convenience more than with backup and protection in mind”.

Outside of his organisation, Wright says he is beginning to see more people looking at object storage solutions, including Amazon S3 and OpenStack Swift derivatives such as SwiftStack, as potential targets for system backups.

Fowler adds: “I am hearing and seeing more people use OneDrive to back up important documents due to the nice integration with Windows 8."

Did Microsoft get something right?

Waters's observations of real-world cloud backup usage is restricted to very small organisations. “They use cloud as an offsite option to store data, for example backing up to USB HDD and storing on Dropbox, OneDrive or Box,” he says.

This is pretty much how cloud backups are currently marketed: the cloud as a disaster recovery solution complementary to a local backup copy. It is an approach that makes good sense.

Brandon Kolybaba sees a very wide array of backup solutions among his client base.

Newer applications coded for the cloud have a unique opportunity to build backups right into the application

“In many cases it's a snapshot in a VM on a nightly basis. In other cases it's an interactive product layered on top of a VM solution such as Veeam or r1soft. In other cases they build the DR backup solution right into the application layer itself (typically with SaaS),” he says.

This brings up an interesting point: newer applications that are coded specifically for the cloud have a unique opportunity to build backups right into the application. Instead of dumping a file to a location and hoping some other application collects it, today's SaaS apps can actually provision storage on the fly and back up around the world at the call of an API.

Pipe of peace

When all your data is in the cloud, how do you deal with restores? For many businesses – particularly smaller ones using upjumped commercial ADSL/Cable connections – that is a lot of data to suck down over a very small (and bandwidth metered) pipe.

I have groused about it on El Reg before, as I feel that I am stuck in the fuzzy grey middle where the cost and lag of restoring from failure is the biggest barrier to adoption. Our experts largely echo these concerns.

“Everyone loves the idea of backups to the cloud. That is, until they try to restore from the cloud,” says Wright

"The very nature of lower-cost cloud-based backup solutions is that they are meant to be asynchronously written to, and the lower priority is always the restore process. It isn't that restores are less important but they are less typical.

“The general experience with large restores from cloud solutions is the same: 'why is it so slow?'"

Mr Wright, it was always thus.

Adam Fowler is a believer in the colocation facility. "The state of broadband in Australia makes restores a timely issue. This is often why renting a rack and having backups localised to the rack location makes more sense for time-critical restores,” he says.

It doesn't hurt that if you need the data you can go to the colo, copy it all to a disk and drive it back to your site. Never underestimate the bandwidth of a truck full of disks.

Vanover calls restores the hard part. He believes that Veeam's ability to do some restores in the cloud, leveraging compute "is a difference maker".

While Veeam can’t do an instant virtual machine recovery in Azure (as one example), it can "extend your backup infrastructure to storage and compute in the cloud and pull back a file from the file server VM that I backed up a while ago, or restore just an email message”.

Granularity, then, is the key. Only pulling back what you need – for example one file from the file server virtual machine instead of the whole virtual machine – may be the key to keeping restore time down in real-world practice.

Kolybaba is unique among the experts polled. Being a cloud provider his organisation does not have the tiny bandwidth problem. “We are only backing up virtualised/cloud environments so we have fat pipes,” he says.

As for ease of use, his techs use purpose-built backup and disaster recovery solutions which make the restore process very simple.

“Typically it's a web-based point-and-click solution,” he says. OpenStack folks have all the fun.

Backup pain relief

Backups have always been a pain. Not everyone has the dedicated staff to properly run their own backups, especially when we start talking about tape archives and so forth.

What options are available to those looking to get the most bang for buck from backups, be it local, cloud or hybrid?

Eric Wright is a proponent of the incremental approach. “The first backup is a full one of course, but incremental backups ensure that we have the continuously adjusting baseline for backups,” he says.

“It is typical to have a three-tier (often known as grandfather-father-son) approach and then we see the ever-popular 30-daily, 12-monthly, seven-annual backup structure typical of financial and healthcare organisations which have more strict retention policies."

Like most of El Reg's storage experts, Wright argues that the world of storage is changing and this is affecting backups.

“With data deduplication appliances and site-to-site replication happening, I see more tier 2 and tier 3 systems staying on disk, but in protected sites. Tier 1 apps are still getting the full off-site treatment in many cases,” he says.

“Hybrid approaches are quickly gaining ground, even if just to test the water. If you ever need an indication of how deeply entrenched in traditional backups many organisations are, just look at the sales of magnetic tape media, which are increasing rather than decreasing."

I agree with Wright that the death of tape is a myth, though I also feel that eventually a lot of the backup duties served by tape today will eventually be the niche that spinning rust arrays end up occupying.

"Hybrid is the sweet spot in the market at the moment,” says Odgers. He thinks customers are moving away from having dedicated staff for every silo (such as backup/recovery/archive) and moving towards a model where they can hire generalists rather than experts to manage IT. As an IT generalist, I have to say I'm a fan of his view of the world.

Waters believes that data should (where possible) be held centrally in an application that provide data retention/archiving capabilities.

“This means whatever backup I have, at whatever time, I have the full copy of all data and do not need to rely on the backup technology to provide versioning of that backup. That way backup products need only worry about backing up a system 'state' for recovery and do not need to worry about what version and so on,” he says.

Home sweet home

I suspect a goodly chunk of the backup industry will see that as fighting words but I must adamantly agree. I still have a number of cases where critical data is backed up by taking the full set, RARing it and letting the backup software archive the rarball. I just can't risk the backup software doing something stupid with the versioning, so I back the whole thing up every night.

Vanover's view of the backup landscape is considered and complex. He thinks that backups are "a blend of a few key things". Among these are self-service recovery, deduplication and an agentless approach to backups. Vanover calls for "awareness" regarding backups.

“A 'backup infrastructure' is an infrastructure,” he says. “Historically, data protection is an afterthought but it has measure points and potential bottlenecks as well." Amen.

There you have it, folks. Cloud backups are real and they are viable – but local backups are likely to still be the thing for some time. It is looking like a hybrid future for many of us.

Your thoughts in the comments, please. ®

More about

TIP US OFF

Send us news


Other stories you might like