With the rise of the “cloud” buzzword in recent years, it has become imperative that it be included in the analysis and planning around the infrastructure decision-making process. This is required in order to accurately portray the Total Cost of Ownership involved with seeking to move from a traditionally Capital Expense endeavor to an Operating Expense model. The overall issue is clouded, no pun intended, by differing purchasing strategies. For instance, a leasing strategy essentially transforms the capital expense of acquiring hardware into an operating expense.
Since “cloud” based infrastructure (IaaS – Infrastructure as a Service) is generally charged on a usage basis, it is sometimes just as or more cost-effective for some businesses where the computing and storage needs tend to be relatively level, to run their own physical, on-premise data center for Production needs. That said, there are opportunities to leverage IaaS and potentially reduce cost in two areas: Disaster Recovery (DR) and Test/Dev.
In the case of maintaining a physical, on-premise DR site, companies incur the cost of compute (servers) and storage with no actual benefit. The hardware sits idle at the DR unless and until it’s needed. A generous Maximum Tolerable Downtimes (MTD – the amount of time a system can be down before it detrimentally impacts the business) and Recovery Time Objectives (RTO – the amount of time required to bring a system back online) is generally referred to as a cold-start plan as servers (compute) need to be attached and/or brought online, virtual machines realigned with them, and, finally, data restored to them. Generally speaking, the lower the MTD and/or RTO, the more expensive DR becomes. IaaS has the potential to alter this calculus so that companies can achieve quicker restorations in the event of a disaster with equal or lower costs.
In the case of Test and Dev infrastructure, companies tend to maintain a similar, albeit somewhat smaller compute and storage footprint in order to support their Production efforts. In theory, Test and Dev are far less utilized than Production, and they would be more suited to be run in a usage based IaaS model.
Finally, there is the Software as a Service (SaaS) category of the “cloud.” This area is best used for commodity type software, such as e-mail. In this instance, companies tend to maintain, for good reasons, a full production copy of e-mail in both Production and DR sites. This is an excellent example, referenced above, of incurring additional cost to maintain a low MTD and RTO for e-mail. Moving e-mail to a SaaS model, such as Office 365, can potentially provide the same benefits at equal or reduced costs.
According to Gartner, the Strategic Planning Assumptions for storage are:
1. By 2018, open-source storage will gain 20% of the market share, up from less than 1% in 2013.
2. By 2019, 25% of all mission-critical workloads will be supported by flash arrays.
3. By 2019, 50% of all SMB workloads will be hosted on cloud infrastructure, platform or SaaS.
In addition, they state that one of the primary reasons for maintaining the status quo is that decision makers are “not bold enough to experiment with new storage technologies or providers” and “business as usual is not going to allow the organization to attain the storage ownership cost reductions needed to avoid becoming a bottleneck in application deployment.” Gartner goes on to indicate that over the next one to two years, IT leaders must use the following guidelines to build a cloud storage strategy:
• Match workloads to the cloud
• Offload those workloads that can be standardized to the cloud
• Keep those workloads that require on-premises customization in-house
I can’t remember where I read it, but when I heard the following, it changed my thinking on cloud quite a bit: An important tenet to remember is that “moving business processes to the cloud is not the goal; using cloud to potentially optimize business outcomes is the goal.”
Therefore, a narrative picture of a future infrastructure would look like this: maintain a physical, on-premise data center for Production needs. However, due to other considerations below, this infrastructure would be smaller and less expensive. Implement Office 365 and move e-mail considerations to the cloud infrastructure, obviating the need to maintain a High Availability portion at an alternate site. Migrate all DR to Disaster Recovery as a Service (DRaaS) where it can be maintained in readily available state. Consider migrating Test/Dev efforts to the cloud where usage can be controlled when Test/Dev resources are not required. The end result would be a smaller Production footprint and a move towards infrastructure as an Operating Expense.
Generally, pure cloud strategies are useful and cost-effective in one of two major categories. The first involves new companies who don’t have the time, knowledge, or resources to build and maintain their own data center. The second involves businesses who have radical spikes in their business models which requires them, in a traditional data center, to size for the peak even though the majority of the time they only require the valley in terms of levels of infrastructure.