The cloud landscape has taken on a bit of the Wild West. Specifically, the mixing and melding of free to low cost cloud storage, apps, etc. Cloud sprawl stems from Shadow IT’s acquiring an overabundance of public cloud services and apps. With company data stored across multiple cloud services, in multiple locations, many IT departments simply can’t keep up with the management and security complexity the phenomena has created.
As little as 3 years ago, cloud sprawl was already becoming an issue. The cloud computing market has matured over the past few years that companies large and small are now acknowledging the importance of cloud governance. This is clearly demonstrated by the tremendous demand of cloud monitoring and optimization services, with the understanding that transparency and control are imperative for optimal cloud cost benefit. Does this mean that cloud sprawl is no longer an issue?
Starting Small Doesn’t Mean No Sprawl
We see new cloud comers every day. They register for AWS cloud services and start purchasing new resources, all in the same breath. The cloud’s flexibility and lure make these new users ignore the mere existence of an invoice, finding that they only wake up when their CFO inquires about the previous month’s outrageous bill. When we discuss the matter with them, however, it is evident that many newcomers are totally aware of the need to retain cost control, which is demonstrated by their interest in planning and monitoring. On the other hand, some say they want to start small, then pay more attention to these details as they grow. The issue there is that the notion of “starting small” actually makes initial costs so low that they will always be ignored. That is, until the CFO calls for a “Cloud Costs” session to discuss the previous month’s exorbitant bill. Unfortunately, this way of thinking can lead to even more difficulties later on.
While the cloud market continues to mature, its associated environments tend to become more and more complex. According to a Gartner report, 50% of large enterprises will have built their own private clouds and have a hybrid environment by 2017. This heterogeneous cloud environment is obviously much more complex than the current state of affairs and could very well lead to cloud sprawl that should have been avoided from the get-go. As discussed in detail in our recent post, “The Compute Sprawl: Is it Really Getting Better?”, the move from virtualization to the cloud didn’t solve the sprawl, on the contrary, it just made it more severe. While the perfect CPU hour might be a goal for achieving high cost-efficiency in the cloud, today we still see a low average of CPU utilization, similar to what we saw in traditional data centers.
So, it’s time to take the bull by the horns with these 3 important steps:
1. Develop a Cloud Consumption Strategy
Developing a customized cloud consumption policy is based on a well defined capacity plan and governance:
- Capacity plan – This is based on your service features, such as your SLA, current utilization patterns, forecasted growth, storage size, and environment configuration. As a result of the factors mentioned above, your capacity plan should be allocated to each individual business unit/project within your company, better ensuring a solid ROI from the cloud.
- Governance – This entails putting the right individuals and methods in place to take ownership over your cloud consumption. This goes for established companies, and startups, alike. Be sure to place someone in charge of planning the consumption as well as monitoring it along the way. This person or team needs to document your cloud footprint, including all of the resources used in your environment, as well as their tags, purpose, types, and location. This will ensure that the appropriate costs are allocated to each cloud resource according to its specific business purpose.
2. Design Cost Aware Architecture
Investing all of your architects’ efforts into designing a great cloud environment may meet your online service’s needs and serve future growth, however, be sure your architects include their deployment design’s initial costs in the equation, as well. Moreover, because of the cloud’s dynamic nature with unlimited resources and flexible provisioning, these estimations need to include a forecast of your cloud costs’ potential growth. We suggest simulating your cloud usage in efforts to better predict your overall capacity for the coming year. This way, you will also be able to commit to the right amount of reserved capacity without the fear of losing money from unused resources.
3. Relentlessly Monitor Your Average CPU Utilization
As mentioned above, the introduction of the cloud led many individuals in the industry to believe that IT utilization would increase to 40-50%. This proved to be wrong, with IT environments only reaching 15-20% CPU usage. Ultimately, cloud consumers’ lack of control over idle capacity and redundant resources is to blame. From the cloud consumer’s perspective, optimal utilization translates into only paying for resources that were actually used. While cloud vendors view utilization as the actual use of physical hardware, as a consumer, you are simply paying for the resources that appear on your cloud console. Compute resources constitute 70% of cloud bills, so the ultimate metric to track is average CPU utilization across an environment. Nonetheless, this is not a simple metric to attain due to the different types and purposes of compute resources (i.e. instances that are solely used for memory and cannot be shut down).
As mentioned above, cloud sprawl is a phenomenon that IT companies and organizations are well aware of today. Yet, the cloud’s flexibility and as-a-service manner lures end users in, encouraging them to consume more and more; from replicating an environment, to performing a quick test, to provisioning additional instances for a new developer. The measures mentioned above should not by any means harm these capabilities. However, if your CFO notifies you of unacceptable cloud costs, the outcome will in turn be limited cloud usage and capacity. The decision to migrate to the cloud needs to be accompanied by these important guidelines, even if you plan on growing at a controlled rate.