Abstract:
Container-based
Internet of Things (IoT) applications in an edge computing environment require
autoscaling to dynamically adapt to fluctuations in IoT device requests.
Although Kubernetes’ horizontal pod autoscaler provides the resource
autoscaling feature by monitoring the resource status of nodes and then making
pod adjustments if necessary, it evenly allocates pods to worker nodes without
considering the imbalance of resource demand between nodes in an edge computing
environment. This paper proposes the traffic-aware horizontal pod autoscaler
(THPA), which operates on top of Kubernetes to enable real-time traffic-aware
resource autoscaling for IoT applications in an edge computing environment.
THPA performs upscaling and downscaling actions based on network traffic
information from nodes to improve the quality of IoT services in the edge
computing infrastructure. Experimental results show that Kubernetes with THPA
improves the average response time and throughput of IoT applications by
approximately 150% compared to Kubernetes with the horizontal pod autoscaler.
This indicates that it is important to provide proper resource scaling
according to the network traffic distribution to maximize IoT applications
performance in an edge computing environment.
Introduction
Edge
computing is a new paradigm that overcomes the inherent limitations of cloud
computing by distributing edge nodes with computing resources closer to IoT
devices. Using containerization, due to the lightness and portability of
containers, it is easy to deploy, install, update, and delete application
services on edge nodes, and various types of IoT services can be provided
simultaneously at each edge node. As such, containerization is widely considered
as the most suitable technology for providing IoT services in edge computing
environments. However, containerization technology is limited to deploying and
managing container-level application services, requiring container
orchestration to monitor and manage resource status through multiple edge nodes
in an edge computing environment
However,
despite the many benefits of Kubernetes, it is still in its infancy in an edge
computing environment. In edge computing infrastructure, requests from devices
are handled by container-based applications on edge nodes, and the traffic load
varies over location and time. Namely, as some nodes are too busy to handle a
large amount of traffic while others are idle, an imbalance of demand occurs
between nodes. In Kubernetes, the kube-proxy balances resource usage between
nodes by sharing the incoming traffic at each node to all pods in the cluster
in a random or roundrobin manner. Because edge nodes are geographically
distributed, there is network delay in their communication, so this kind of
redirection offered by the kube-proxy can increase the response time of
applications. Therefore, it is necessary to allocate more or terminate
redundant computational resources according to network traffic at each node to maximize
the amount of locally handled traffic while minimizing network delay by
minimizing the number of requests handled by pods on remote nodes.
Nevertheless, when the resource demands of the applications change, Kubernetes’
HPA (KHPA) only tries to evenly distribute new pods to nodes or terminate
redundant pods on nodes based on pod status without considering the network
delay between edge nodes and the volume of network traffic accessing them in
real time. This limitation of KHPA can result in the degradation of the quality
of service and overall throughput of the system.
To solve the
aforementioned problem of KHPA in a Kubernetes-based edge computing
infrastructure, this paper proposes traffic-aware HPA (THPA), which operates on
top of Kubernetes to provide dynamic resource autoscaling by considering the
IoT service demand at each edge node. Specifically, in an upscaling event, THPA
allocates a number of additional pods proportional to the distribution of
network traffic accessing the nodes, whereas in a downscaling event, it
terminates the pods in the node with low demand. Experimental evaluations prove
that THPA significantly improves the average response time and throughput by
maximizing the amount of traffic handled locally and avoiding the round-trip delay
from redirection between edge nodes in an edge computing environment.
Base Paper - https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9709810
0 Comments