Everyone's going to the cloud, it seems, but when it comes to data storage, companies need to exercise particular caution. You may be attracted to a potential low cost per GB, but when you consider the investments that may need to be made to sustain a cloud storage and backup strategy, you could find that your "fully loaded" per GB costs end up being a lot higher than expected.
Cloud is everywhere -- there is no escaping it! This is actually a good thing for companies having to deal with burgeoning storage and backup costs with no relief in sight. Internet-based cloud services provide a viable alternative for companies looking for an additional location to store their data.
The reason I say additional is I believe that for a vast majority of companies, the public cloud is not yet ripe enough to be made the primary location for storing their data. For the lucky few who have made the switch to public or private clouds for data storage, this article may validate some of their assumptions and perhaps offer a different perspective on the risks leading up to that decision.
For those who are still in the evaluation stage, I would like to offer a word of caution: This article presents some high-level assumptions that may not apply to your environment, so full due diligence should be a key part of this process.
So, let us take a look at how you can avail yourself of storage and backup services in the cloud while ensuring that your data is safe and service levels are maintained.
1. Cloud Is Proprietary, So Due Diligence Is a Must
No matter how you look at it, almost every kind of cloud service rests on a proprietary platform. This means that once you sign up for that cloud service, you have conducted some form of vendor lock-in.
Considering that this is no different in the storage and backup services world, the fact that cloud is proprietary should not be that worrisome. However, keep in mind that the time and energy it takes to put data in the cloud changes this perception. Moving it from one location to another will take twice as much as it takes to put it in.
So, before you start dumping your precious data in the cloud, you should do a full evaluation of the cloud service provider's offering. Examine things like how the proprietary nature of its services will impact your overall storage and backup environment. Examine the platform and interoperability support. For example, some service providers only support Windows and Linux platforms. If you are a Unix shop, this could present a challenge. Next is the question of how your storage and backup software will interface with this service.
2. Send a Tertiary Copy of Your Data
A strong recommendation for those starting out anew in the cloud is to consider putting their tertiary copies of data in the cloud. What this means is from within their backup software, they can clone or duplicate their backup data to the cloud; or, using cloud for storage services, they can use replication or copy software to create a replica of their data.
Backup vendors are beginning to offer integrated APIs for cloud services, but until your installed stack is fully supported, you may have to rely on traditional methods to move data in and out of the cloud -- for example, a tertiary asynchronous mirror for storage or a backup to disk locally followed by cloning of that data to the cloud.
Once the data is in the cloud, evaluate how that data can be accessed. Again, the question of relying on traditional access mechanisms comes to the fore.
3. Remember, It Is the Size of Your Network Bandwidth - not Theirs
A lot of customers lose sight of the fact that the moment you talk about an Internet-based cloud service, you are leaving the comfort of your own network and venturing out to the Internet to access data. It is important to investigate how this access will play a role in how you access or protect your data.
Network latency, which in your local environment is generally minimal to nonexistent, can suddenly become a huge problem. Imagine going from access times of less than 10 milliseconds to something over 20 or 30 milliseconds? Yes, it is important that the cloud provider has a fast pipe to allow customers to access their service, but what is more important is the size your network pipe to access that service -- whatever route it takes to get there.
Even more importantly, this network will more than likely be a shared service (unless you are willing to invest in a dedicated network pipe for cloud services), which means that the cloud access traffic will struggle alongside all kinds of Internet traffic. This could become a challenge during business hours when there is heavy Internet usage.
Storage traffic is not network conservative (certainly not going by the race to the top in speeds and feeds of Fibre Channel, Gigabit Ethernet, etc.) and unless your copy in and out of the cloud has some kind of asynchronous features built into it, it has the potential to impact your primary environment. For example, your backup environment may run off time cycles or resources cloning data to the cloud, thus impacting your primary backups.
4. Private vs. Public Cloud Storage Not Settled
This topic is not fully baked yet, so expect this to continue evolving in the foreseeable future. The question here is no different than the paranoia surrounding traditional data, storage and backup security. If the idea of someone snooping on your data concerns you, then moving to the cloud is something you need to think long and hard about.
Luckily, private cloud services offer some security in this area. For the purpose of this article, let us consider a "private cloud" as "private access to a public cloud" and not "building out a private cloud" -- the latter is more for enterprise IT departments that can afford to or want to offer a cloud service to their own customers -- and, more importantly, on a scale that warrants such an investment.
For most companies, the former definition of "private cloud" means that the service is analogous to virtual private networks -- a service that uses the Internet to access company networks, but is secure and virtually private. However, there should be (and companies should demand) one big distinction -- how segregated or logically isolated your company's data is at the cloud service provider's location.
From a storage-and-backup-software perspective, this should not matter and will be no different than a shared storage environment. A storage array can be home to many applications running on multiple different servers, all accessing the same set of physical spindles, but if there are any regulatory requirements that require spindle separation, you may have to demand isolation from the service provider as well.
5. Long-Distance Network File Share Challenges
A lot of companies are using cloud services to allow their users access to a network home directory or file share via the cloud. Great concept; however, there could be challenges -- not the least of which is a quadruple increase in the amount of traffic going out to the Internet.
This is on top of the increase in latency introduced due to congestion. Normally, such traffic would have remained local to the company network. Also, any issues in access times can be quickly resolved by internal network, storage and systems teams.
In the case of cloud, network issues will mean involving the cloud service provider and potentially the Internet service provider. This leads to many hours of troubleshooting that could eventually result in many unsatisfied users. Scale is important here, because if you are planning to roll this out for the entire organization, you may want to first examine the impact. A staggered rollout may be necessary.
6. The Cost per GB Question
A lot of customers ask me about how moving to the cloud impacts their storage and backup cost per gigabyte.
If you examine all of the topics above and think about the potential investments that need to be made to sustain a cloud storage and backup strategy, this is the answer: Moving to the cloud has the potential to have a significant impact.
So, even though at the periphery, your "cloud" cost per GB is low and you are ecstatic at being able to move some of your CapEx costs to OpEx, when all is said and done, your "fully loaded" per GB costs could be a lot higher. Think about it.