As Identiq’s Director of Cloud Operations, Oded Avissar leads a DevOps team that is responsible for ensuring that customer installations go smoothly and for managing the Identiq P2P network — all while striving for high uptime and keeping costs down.
The best way for Identiq to optimize cloud costs was through the commitment-based discounts offered by AWS like Reserved Instances or Savings Plans, which are offered on 1- or 3-year terms. However, as a scaling startup with a rapidly-changing product, committing to even one year of EC2 usage was a challenge for Oded. To commit, he would need to coordinate with various engineering teams to scope their resource requirements for the coming year – a difficult exercise in its own right.
“If I optimize my code tomorrow and require fewer resources…I don’t want to be stuck with an over-commit.”
Scoping such requirements goes beyond simply quantifying needed compute capacity. The team would also need to determine which machine types they’d be using and in what regions around the globe. Adding to the complexity is the fact that those requirements can change over time; customers can come and go and thus alter the regions in which Oded’s team needed compute capacity. Similarly, machine types can be phased out as software changes, and projects can be ended ahead of schedule and their associated VMs turned off. In sum, forecasting their needs with specificity and accuracy made compute commitments more of an art than a science. “If I optimize my code tomorrow and require fewer resources…I don’t want to be stuck with an over-commit.”
Oded and his team also had to consider their finance team – an important internal stakeholder that prefers to work in fiscal-year cycles. If Oded’s team discovered they needed an RI in the middle of the year, that commitment would be spread out over two fiscal years, adding complexity for the finance team.
Potential savings from AWS Savings Plans were trumped by the risks of not being able to meet commitments and creating additional work for the DevOps and Finance teams. As a result, the cloud operations team wanted to wait for their infrastructure’s compute requirements to stabilize before purchasing AWS Reserved Instances (RIs) or Savings Plans (SPs). In order to wait, Identiq opted to pay the premium rate of on-demand workloads for all their EC2 compute.
Given these challenges, coupled with the desire to optimize EC2 unit costs, Oded sought a way to realize the benefits of commitment discounts without adding the risk or unused overhead to his company.