Beware of Wrongsizing: Avoid These Top Five Rightsizing Mistakes

Not all cloud rightsizing recommendations are built equally. Bad decisions (based off poor recommendations) could leave your infrastructure “wrongsized.” Learn about the mistakes and their costs: lost efficiency, savings and wasted time for your engineering teams.
Every cloud practitioner’s goal is to deploy resources that are well matched to their corresponding workload. For this task of rightsizing, it makes a lot of sense to employ a purpose-built tool to deliver recommendations which allow you to avoid the guessing game.

However, not all cloud rightsizing recommendations are built equally. Bad decisions (based off poor recommendations) could leave your infrastructure “wrongsized.” The costs: lost efficiency, savings and wasted time for your engineering teams. Even worse, you could end up compromising the performance of business-critical services.

If you suspect your current tool might contain invalid recommendations, it’s time to consider a better rightsizing solution. To make sure your rightsizing tool doesn’t leave you wrongsized, be aware of these five mistakes.

Mistake 1: Relying on Recommendations That Don’t Account for Spikes in Utilization

Taking into account shifts and spikes in utilization, and then applying a statistical model to enumerate the best-matched AWS resource types isn’t a simple matter. Most tools out there that we’ve seen use some kind of “rule of thumb” based only on utilization averages over a time series and the tightest-fitting resource type is recommended.

Using only a utilization average can lead to two things. The most common scenario is that the engineer realizes the recommendation is completely bogus and the cost saving initiative ends. They’ve wasted a bunch of time and their deployment is still costing more than it should.

The other scenario is far worse. The less-experienced engineer may go ahead and take the recommended action. During quiet times—even average times—the new resource configuration will likely hold up fine. But when a regular spike or shift occurs, things begin to seriously unravel. Demands on the resource push it beyond what it’s capable of: performance starts to degrade and it’s possible to see some sort of outage. Any potential cost saving is vastly outweighed by the risk put to the business.

Mistake 2: Failing to Rightsize Beyond EC2

Everyone loves focusing their energy on EC2, but rightsizing tools that only recommend changes within this AWS service are only solving for one part of a bigger cloud cost picture. Here at Cloudability, we see utilization issues across the board, but one particular spot worth mentioning is RDS. This is especially true as we see the use of RDS Aurora growing. If your rightsizing tool can’t recommend across non-EC2 services such as RDS you’re leaving a whole bunch of potential savings on the table.

Mistake 3: Not Addressing Your Resource “Shape”

Rightsizing decisions shouldn’t be stuck within the one EC2 family. AWS has numerous, specialized EC2 families and you should choose one that matches the resourcing shape of your workload. For example, the memory on a particular R4 instance may be drastically over provisioned, but maybe CPU is about right – this is an opportunity to get your shape correct by keeping compute consistent while at the same time reducing your memory. In this scenario a sophisticated tool will save you significant money by recommending you go from say an r4.large to c4.large. Again, if your tool won’t recommend across families you are leaving serious money on the table.

Pro-tip: Check out our resources on Choosing the Right EC2 Instances to learn more about the different families.

Mistake 4: Not Simulating Performance Before Rightsizing

Using rightsizing recommendations that don’t forecast how those changes can affect your infrastructure can lead to hesitant and sub-optimal decisions. It’s highly preferable that your rightsizing engine give multiple recommendations and a means to visually validate each one. That way you can inspect the risk or probability that there could be clipping and take that into account when comparing to the savings you’ll make. Without this visualization technique you risk being too conservative (limited savings) or too aggressive (performance hit/outage).

cldy emerge 2017 06 clipping 08 2815 - Beware of Wrongsizing: Avoid These Top Five Rightsizing Mistakes - ApptioYour teams might be concerned with “clipping” and will make sub-optimal decisions. But with the means to get a forecast of how the recommendation affects their infrastructure, they can make better choices to optimize their cloud.

Mistake 5: Hesitating Due To Reserved Instance Uncertainty

One common thing we hear at Cloudability is, “I don’t want to rightsize at all because it might affect my RI coverage and cause waste.” There was a time when this certainly was a consideration, and caution was required. But the good news is that AWS has introduced tons of options which give far more flexibility and allow us to rightsize with courage. With an intelligent RI planner that is built around smart ISF recommendations and can work across Convertible RIs, you’ll be able to rightsize with confidence.

How Rightsizing Should Work

Now that you know what to look for in order to avoid wrongsizing your infrastructure, let’s talk about how to do it right. Any effective rightsizing tool should:

  • Allow for any spikes in utilization when giving a recommendation (key resource metrics like minimums and maximums of utilization metrics can add a lot of context)
  • Rightsize across all the key AWS products (not just EC2)
  • Rightsize into AWS resources that best fit your workload (go beyond a sole EC2 family)
  • Provide a means to visually validate multiple recommendations
  • Provide rightsizing recommendations alongside Convertible Reserved Instance options or ISF options

This holistic perspective on rightsizing lets you make the right technology decisions to operate a leaner cloud, cut down on waste and reduce operational costs. For a closer look at a solution that won’t leave your infrastructure wrongsized, we invite you to see our rightsizing tool and RI planner at work with your own cloud cost and usage data.

Article Contents

Categories

Tags

Additional Resources