A company is observing that their Google Cloud spend on Compute Engine instances is consistently high, even during off-peak hours. Many of these instances run stateless web applications that experience predictable daily traffic patterns. They have a stable baseline workload but also significant spikes. The current strategy uses on-demand VMs. What is the most effective cost optimization strategy for this scenario?
- AUse Sole-Tenant Nodes to ensure dedicated hardware and reduce licensing costs.
- BConvert all instances to Preemptible VMs (PVMs) and implement a retry mechanism.
- CMigrate all stateless applications to Cloud Functions to leverage serverless cost models.
- DPurchase Committed Use Discounts (CUDs) for a portion of the stable baseline workload and implement autoscaling for the fluctuating part.
Show answer & explanationAnswer & explanation
Correct answer: D. Purchase Committed Use Discounts (CUDs) for a portion of the stable baseline workload and implement autoscaling for the fluctuating part.
For a workload with a stable baseline and predictable spikes, a hybrid approach combining Committed Use Discounts (CUDs) for the baseline and autoscaling for the variable portion is the most cost-effective. CUDs provide significant discounts for predictable usage, while autoscaling ensures resources scale up only when needed for spikes, avoiding over-provisioning and high on-demand costs.
Why the other options are wrong
- A. Sole-Tenant Nodes are for specific isolation or licensing requirements and are generally more expensive, not a cost-optimization strategy for high Compute Engine spend on standard stateless web apps.
- B. Preemptible VMs are for fault-tolerant, interruptible workloads. Stateless web applications, while stateless, typically require continuous availability during active hours, making PVMs unsuitable as a primary compute solution for critical web traffic.
- C. While Cloud Functions can be cost-effective, migrating 'all' stateless web applications might require significant refactoring and may not be suitable for all application types or existing instance configurations. It's also an architectural change, not just an optimization of current compute spend.
Committed Use Discounts (CUDs)
Deep discounts on Google Cloud resources (like Compute Engine VMs) in exchange for committing to a specific level of resource usage for a 1-year or 3-year term.
- Offer significant savings (up to 70%+ for 3-year commitments).
- Apply automatically to eligible usage in the billing account.
- Best for predictable, stable workloads.
- Can be combined with autoscaling for variable workloads.
Memory trick: Commit for baseline, auto-scale for bursts, save big!