kepler: Kubelet update is blocked by network IO?

When I try to add benchmark for updateKubeletMetrics I found it seems a blocked network IO?

Here is the tracing stack:

https://github.com/sustainable-computing-io/kepler/blob/3f117b3c8b198806a7f26cc0e46bb18830bb0789/pkg/collector/container_cgroup_collector.go#L45

func GetContainerMetrics() (containerCPU, containerMem map[string]float64, nodeCPU, nodeMem float64, retErr error) {
	return podLister.ListMetrics()
}

// ListMetrics accesses Kubelet's metrics and obtain pods and node metrics
func (k *KubeletPodLister) ListMetrics() (containerCPU, containerMem map[string]float64, nodeCPU, nodeMem float64, retErr error) {
	resp, err := httpGet(metricsURL)
...

hence each time for updateKubeletMetrics, before it goes to loop containers/pods, there is a IO at networking to get and convert kubelet. Usually we thought IO is blocking, as the major time cost, and meaning less for benchmarking based on container. Suggestion as decouple this network IO from code stack.

and btw, why we rebuild constants below for each container?

	cpuMetricName := collector_metric.AvailableKubeletMetrics[0]
	memMetricName := collector_metric.AvailableKubeletMetrics[1]