Just ran a quick test on our dev cluster after the update. The speed boost is real! Scanned 200 container images in ~8 minutes instead of ~9.5.
Our setup is mostly EKS with a mix of on-demand and spot nodes. Noticed a few things:
* CPU usage on the scanner pods dropped slightly (avg 0.85 vCPU vs 1.0 before).
* No change in memory footprint.
* Our Terraform module for deploying the scanner needed zero tweaks.
Anyone else seeing similar gains? Especially curious about large, multi-region deployments. Could this translate to smaller scanner auto-scaling groups? 🤔
#savings
>CPU usage on the scanner pods dropped slightly
That's the key metric for cost. Lower CPU at same throughput means you can scale down the node pool.
We run GKE, saw similar speed gains. But watch out for bursty scanning loads - your HPA might react slower to the reduced CPU consumption. Had to adjust our target utilization.
Optimize or die.
You make a good point about HPA adjustments. We've seen the same need to recalibrate after performance improvements, not just for CPU but sometimes for custom metrics tied to scan completion time.
Our team ran a quick analysis correlating HPA response lag with cost savings. While you can reduce node pool size, the slower reaction to bursty loads meant we had to increase our buffer pods, which partially offset the savings. The tradeoff wasn't linear.
What target utilization did you settle on after your adjustment? We're still tuning ours and found that a simple CPU threshold change wasn't enough, we had to add a secondary metric based on the scan queue length to keep up with sudden spikes.
Data > opinions