Having your data hosted in the cloud is enticing, but have you assessed the possible risks?
Cloud storage is fast, easy, scalable and safe. Or is it? Hyperscale clouds such as AWS, Azure and Google Cloud Platform are offering storage as a utility ready to serve businesses at the click of button. But that doesn’t mean they are the most appropriate solution for all workloads, or that they don’t present any business risks.
We take a look at five things you need to consider before diving head first into a cloud storage solution, and paying for it later. Let’s focus on storage and the specifics of housing data in the cloud. Cloud storage in this context means any combination of block (typical server storage) or object storage, which can be used in a transactional way.
1. Cost of performance. Cost is always important and tends to be front-and-centre when deploying a cloud solution, but when it comes to storage the cost of performance matters a lot. Clouds use shared network and storage infrastructure, and while the move to SSD technology has improved disk performance, high latency and capped network throughput can impact your application and, ultimately, end-user experience. Ensure you investigate the risk of sacrificing storage performance for convenience.
2. Flexibility and management. By their nature cloud storage services are designed to service millions of customers in a consistent and scalable manner. This might be fine for some of your workloads, but you can easily run into challenges with what the hosted service allows you perform, and what your business demands. Compare the features you get with a dedicated storage system (or software-defined storage) with those of the clouds and assess the levels of flexibility and management you get.
3. Data protection and retention. Backups, data protection and data retention are all crucial processes for most enterprises and cloud storage services won’t magically perform them for you. For better or worse, with infrastructure you control you can determine how long files and other data is retained and the extent to which it is backed up. While the cloud does a good job with system redundancy, in most cases backups, disaster recovery and retention of deleted data must be stipulated by the customer. For example, AWS has its Glacier service for retention purposes. In the case of DR, don’t get caught out with a cloud outage. Ensure you have a good process for dealing with business continuity before the move.
4. Skills and lock-in. To get the most out of cloud storage you will need new skills and processes to adapt to their way of doing things. Review the skills investments required and whether this change will translate to high ongoing costs for a discrete capability. Another challenge is data portability and standards. Storage clouds have their own APIs and retrieval policies (and charges), which might not be easy to migrate away from if there is a business or technology case for a move.
5. Scalability value. Clouds make a lot of noise about their scalability prowess, but customers need to be careful the storage value proposition doesn’t diminish at scale. As your storage requirements grow the dynamic of the price-to-capacity equation will also change. For small workloads cloud looks great, but for high amounts of storage cloud can be expensive. It’s not unusual for companies to use their own (managed in-house or by an MSP) storage equipment for large processing requirements and come way out in front in terms of scalability and price compared with a public cloud. At the extreme end, Dropbox famously moved off AWS to its own storage infrastructure.
The cloud offers many options for storage, but you need to properly assess the long-term implications for your business. If you are heading into the cloud, be sure to fully understand the retention and data protection requirements.
The ideal architecture will allow you to process your data in the cloud or on-prem without a huge time or cost burden. Storage is one of the best uses cases for a hybrid cloud strategy so assess all your options to avoid being tied down to one type of infrastructure model.