If I had a nickel for every time I heard an IT architect say cloud storage costs less than local storage, I would have a big pile of nickels. But I would still not have as big a pile of nickels as enterprises pay to cloud providers every month for poorly planned storage agreements.
The idea of cheap online storage with no upload fees is enticing, especially when you are comparing it to maintaining underutilized hardware in your own data center. Gotchas occur when you accidentally select the wrong type of storage plan or if you fail to anticipate the volume of data you will need to extract from the cloud into your systems.
Let’s look at three factors that organizations must consider when shifting workloads to the cloud: volume of data, cost of storage and transport, and disaster recovery plans.
An organization deciding whether to migrate workloads from traditional data center storage to the cloud needs an annualized cost model that takes into account four points: 1) transition costs to migrate data from on premises to a cloud provider, 2) the volume of data the organization generates and exports to systems outside of the cloud, 3) the additional cost of backup of data internal to the cloud, and 4) the cost of updating and testing disaster recovery models.
In most cases, an organization uses 30 percent or less of its total storage capacity. To conduct a true comparison between on-premise data storage and cloud storage, a company must consider not only used and available host-based storage but also the unallocated space on storage devices themselves; unused storage is tantamount to wasted capital.
For example, a compute-intensive application that carries out big-data crunching is one that is well-suited to the cloud computing cost model. But a data-intensive application that uses the cloud as primary storage for jobs such as data warehousing should be considered more carefully. If the data download is just an example or summary of the total data set, result of computation or metadata, then it may make sense to leverage cloud storage. But if the data feeds on-prem data marts, it may not be cost-effective to use the cloud.
Using the cloud to back up data is somewhat cheaper than using it as primary storage for active data. Restores are typically infrequent, so exporting data out of the cloud on occasion doesn’t create large unexpected surges in monthly billing. Determine what part of the environment stores data that rarely needs access, i.e. data needed for legal purposes like email archives or call recordings. These environments can use cold storage from a cloud provider at roughly 25 percent of the cost of other cloud storage.
Lastly, don’t forget to consider the best practice design considerations that are used for mirrored or replicated storage. By default, cloud providers include a secondary copy of all your data with basic storage, which customers don’t always realize is included. In some cases, paying for backup service is unnecessary based on the default capabilities of the cloud provider.
Organizations that store their data in a public cloud often overlook one important point: moving data into the cloud has no transfer cost per unit, but moving data out is billed per gigabyte. Organizations need to fully understand their input/output operations per second (IOPS) so they can understand the financial impact of cloud data storage.
If one of the business objectives for moving storage to the cloud is to reduce costs, an organization must carefully consider application requirements and the expense of transferring data out of the cloud. Storage access speeds are impacted by network bandwidth, throughput and latency. Accessing data from on-prem storage solutions is extremely fast because these solutions are based on LAN networks, while cloud storage transfer speeds in/out of a cloud provider are at the mercy of WAN connectivity. If compute loads are maintained in private data centers with some or all storage moved to the cloud, application performance requirements should be considered.
If an application merely takes in data, crunches a result and delivers a consolidated output, there is likely to be insignificant impact to recurring bills. But, more often than not, clients are caught off guard by the uptick in network costs when a large volume of data is read back from cloud storage after upload. This doesn’t necessarily mean the cloud is not cost-effective, but it may not lead to predictable monthly bills within a narrow range of variance.
In the end, an enterprise can save substantially when leveraging cloud storage. Application design and deployment models are important considerations, especially if there’s “chatter” between workloads.
Every great disaster recovery (DR) plan is built around two key concepts: recovery time objective (RTO) and recovery point objective (RPO). If your RTO is six hours, your business can survive for six hours without access to key systems and data. The shorter the RTO, the more expensive the DR plan. If your data backup to cloud is real-time – or even near real-time – you could consider the RPO to be seconds or minutes. But how cost prohibitive does that make your RTO? Can you get the data downloaded to the DR site fast enough?
The answer lies in how much bandwidth and end-to-end throughput – from data storage device to data storage device – is available to the DR site. The data must pass through a lot of infrastructure to get from storage in Amazon Web Services (AWS) to a hard drive in your DR site.
Assume you have just one terabyte of data in cloud storage. Your primary data center goes offline due to a natural disaster and you need to transfer that one terabyte of data to the DR site. Luckily, backup to the cloud is nearly real-time, so virtually no data is lost and your 10-minute RPO is achievable. You call for DR plan activation, and, within the hour, systems are ready to receive data from cloud storage over the 100-Mbps connection into the DR site.
That one terabyte of data will take more than 23 hours to transfer. If your DR site has a one-Gbps connection, download time is reduced to just over two hours. This meets your RTO, but it could be so expensive to maintain the DR infrastructure around low-cost cloud storage that the solution becomes untenable.
Cloud storage offers exciting opportunities to reduce costs and add flexibility that is difficult to achieve via your own data center. But proper planning and a candid assessment of your application data and storage needs – including whether the cloud can meet your DR requirements – is crucial to any evaluation of cloud storage for your business.
ISG helps companies evaluate their storage needs and navigate the changing landscape of business technology. Contact us to discuss how we can help you.
About the author
Spencer Suderman works with enterprises to achieve IT operational efficiency through increased automation, lessening the impacts of service degradation and increasing the productivity of human resources. He is experienced in improving IT as a strategic asset through network architecture, automation, workflow reengineering and supplier management. As a Principal Consultant in ISG’s Network & Software Advisory Practice, Spencer is focused on helping clients identify and capture cost and value optimization opportunities in their enterprise, ERP, application and subscription-based software and IT hardware. Before joining ISG, Spencer led strategic efforts at a well-known theme park and resort operator to increase the scope of network automation to support digital transformation, enable planning for software-defined networks and network functions virtualization. Additional workstreams to improve the development of network infrastructure standards led to reduced mean time to repair outages and service degradations.