Beyond CPU and Memory: Autoscaling with Custom Metrics
Feb 6, 2026
Most teams autoscale Kubernetes workloads based on CPU and memory — but those metrics rarely reflect what actually matters to the business.
Salman Iqbal, an instructor at LearnKube and a platform engineering consultant, explains why he prefers custom metrics such as requests per minute and queue depth to inform scaling decisions.
In this interview: • Why CPU and memory are poor autoscaling signals for most applications • How KEDA enables event-driven autoscaling with business-relevant metrics • Why scale to zero matters for both cost optimization and reducing your carbon footprint
The takeaway: autoscaling should start with what the business needs, not what's easiest to measure.
Relevant links
Transcription
Bart Farrell: Who are you? What's your role and where do you work?
Salman Iqbal: Hi, my name is Salman Iqbal. I'm an instructor at LearnKube. And I also work as a DevOps and platform engineering consultant for eSynergy.
Bart Farrell: Now, what are three emerging Kubernetes tools that you're keeping an eye on?
Salman Iqbal: I'm keeping an eye on Crossplane, KEDA for autoscaling, doing a lot of work on KEDA. And also for monitoring a new tool. New tools are coming out, keeping an eye on monitoring tools as well.
Bart Farrell: Jorge argued that Kubernetes-based scaling solutions like KEDA offer advantages over traditional monitoring tools like Prometheus, especially regarding responsiveness. How do you autoscale workloads in Kubernetes? What metrics do you use? And are there any tips and tricks that people should know?
Salman Iqbal: For me, the scaling should be based on what the business requirements are. I don't really tend to go with memory or CPU. I go with the requests per minute for an application or messages staying in the queue. I'm looking at what does the business want to scale up. I'm using those custom metrics like messages in queues or number of requests coming to an application.
Bart Farrell: Our guest Brian believes one of KEDA's most distinctive features is its ability to scale deployments down to zero replicas, which is particularly valuable for cost optimization development environments. How important is scale to zero capability in your experience?
Salman Iqbal: Very important. As you mentioned, the costs, we need to keep the costs down. And more importantly, it's good for the environment. We need to keep our carbon footprint down. I definitely agree that scale to zero should always be used.
Bart Farrell: What's next for you?
Salman Iqbal: At KubeCon? Just go around and interview people and ask questions and make some clips for Daniele.
Bart Farrell: How can people get in touch with you?
Salman Iqbal: They can find me on LinkedIn at Salman Iqbal. And also I have a YouTube channel so you can subscribe to Salman Iqbal.
