Yesterday at Google Next, Urs Hölzle quoted a great stat by RightScale – users waste 45% of cloud resources that they buy. While this number is not too far from what typically happens in traditional data centers, which operate at 20-30% of capacity on average, cloud promises a pay-per-use model. You provision and pay for only what you utilize. This promise leaves the user with the impression that they will effectively achieve 100% value—reducing costs significantly compared to an inefficient data center. And while Google’s new committed pricing model tries to lessen the impact, it does not address the root cause of the problem.
So, what is the underlying problem?
The real issue is that many organizations still use cloud like they use data centers, managing their resources in the same manner. What do I mean by that? Let me give you some examples:
- Virtual machines (VMs) spun up and maintained as persistent resources even though nothing is provisioned on them yet.
- VMs spun up for temporary tasks but forgotten to be spun down. An example: developers launching myriad test workloads, inadvertently causing VM and cost sprawl. Then to make matters worse, it later becomes difficult to identify which VMs to spin down and which to keep active.
- Workloads not using VMs at anywhere close to full capacity. This is a right-sizing problem. Users typically over provision where instead a lower cost instance type might be good enough for that workload.
And let’s not forget about managing the application lifecycle.
Traditional approaches involve writing custom scripts, creating cloud-specific VM images, and often having to re-write portions of the application to conform with the specific APIs and services of the cloud provider. One of our large customers recently told us that the typical process to onboard/change a single component of an app requires them to write about 1200 lines of code, and anytime the app or cloud changes, they have to re-write about 20% of the code/scripts. This all adds up to rising cloud costs.
Offers from cloud providers, such as discounted prices for reserved capacity, do not really help solve this problem. In fact, they only exacerbate the problem by forcing organizations to make up-front decisions and continue using the cloud more and more like a data center. And most importantly, what cloud providers often don’t offer users are the intelligent orchestration and management tools they need to solve this problem.
The cost promise of cloud will always remain elusive if organizations are not able to realize the full benefits of cloud’s flexible model. And when you consider that – according to IDC – 85% of organizations will use multiple clouds, this issue is only getting worse!
Automate and Save
Your environment is likely too complex to manage without cloud intelligent tools to help you optimize your spend—be it a single cloud or across multiple clouds.
Cisco CloudCenter can be your answer. Cisco CloudCenter is helping local and global organizations alike get the most of their cloud resources. We do this in several ways to help optimize your cloud spend. For example:
- Policies like auto-aging that shut down workloads automatically after a certain time. These policies are enforced by administrators and lower your cloud costs. For example, a DevOps administrator may enforce an auto-aging policy on all workloads spun up by a development team, shutting down the tagged workloads within a certain timeframe, preventing VM sprawl.
- Auto-scaling or utilization-based scale-down. Based on actual utilization, the number of nodes in a cluster can be scaled down to prevent the use of unnecessary VMs.
- Benchmarking. With the ability to compare different clouds or different instance types on the same cloud, users can see where the best price-performance is achieved for an application deployment time to avoid running on sub-optimal instance types and clouds.
- Most importantly, an app-centric approach. Cloud VMs are only spun-up when an application is deployed, need not be pre-provisioned, and are automatically de-provisioned when workloads are terminated.
The promise of cloud is real, and you can get what you’re paying for. By providing you the cloud intelligence you need for your hybrid IT environment, Cisco CloudCenter enables you to achieve cloud’s true pay-per-use promise. Let’s us show you how.
Good points about resource utilization.
Sounds like a great idea Fabio, and perhaps some change management to supplement.
Very interesting point that users waste 45% of cloud resources that they buy and how the problem can be solved.