Jul 7, 2021. By S V Aditya
Edge computing is growing at a tremendous pace - a 29% CAGR through 2020 - with new applications developed to leverage the underused capabilities of edge devices. Instead of using data centers, these devices use local hardware for capabilities like data preprocessing and decision-making. Consequently, enterprises can deploy smart applications that do not rely on their cloud-based servers heavily. In turn, it improves customer experience as these applications can deliver higher quality services. These edge nodes can be any platform - from mobile devices to full-fledged data centers. At the same time, more and more organizations are developing applications as microservices deployed into containers. Enterprises can use such architectures to improve the degree of monitoring, simplify bug fixes and releases, and manage resources better with container orchestration tools like Kubernetes. While Kubernetes is a great tool for cloud computing management, it requires improvement to service edge computing applications. This is where tools like KubeEdge come in. KubeEdge is an edge computing framework built on top of Kubernetes specifically designed to handle edge computing requirements. It handles resource management and runtime capabilities for container-based applications and simplifies large-scale deployment in edge computing. It is a powerful tool for handling time-critical and decision-critical use cases which should not have to refer to the cloud for processing. Such use cases can be defect inspection in the assembly line or capturing the number plates of a vehicle in a speed trap.
While KubeEdge can power incredible applications in the edge nodes, enterprises rely on more than edge devices - There are centralized cloud servers controlling these devices. There are security protocols associated with these devices. There's network traffic that has to be monitored and there will be many incidents and security risks. We believe AIOps can fill these gaps and create a dynamic system that can respond to complete enterprise needs.
To understand this, let's examine the similarities and the differences between AIOps and KubeEdge. Both provide a solution to collect metrics, logs, and traces from edge devices. While KubeEdge can show you metrics and tell you what is happening, an AIOps in its Observe element can tell you why this is happening and what's the root cause of the issue with Incident Recognition. It can go beyond the scope of an edge device and look at all CIs (whether on edge, cloud, or any server) and tell you the issue in the context of a business service. KubeEdge can restart a containerized service in case of failure. AIOps can tell you why it failed in the first place. An ITOps team using both can have reliable serviceability by restarting failed services in seconds and fixing future defects with powerful diagnoses.
Then, there's orchestration. KubeEdge and AIOps solutions both use a lightweight agent that can power workflows and automate orchestration. In case of a break-in network connection between the edge node and cloud server, KubeEdge continues to operate the device with preconfigured workflows and updates the metadata once the connection is re-established. In the case of AIOps, an AIOps powered agent can react to a change in the situation with AI-based decision-making capability and service more unique requirements. For context, a KubeEdge agent can continue to service an application in case of network failure and then synchronize metadata later. It does this by using pre-built configuration files. An AIOps agent can update these configuration files with AI-powered workflows that react to changes in the system. In effect, it can give the same instructions the cloud server would do even when they are not connected and acts as a proxy to the cloud server. Edge device metadata is also regularly updated whenever an edge device reconnects to the network. AIOps can ensure that this metadata is preserved and is not accessed by unauthorized entities with its workflows.
KubeEdge is still very new with the v1.0 release in 2019. Most ITOps teams would not have the required experience in KubeEdge(or Kubernetes, for that matter) and have to build it from the ground up in a small team. This team then becomes the sole vanguard that can handle all KubeEdge-related issues. As enterprise ITOps teams have multiple responsibilities and a large scope of work, this is a recipe for overburdened and stressed teams. AIOps can help these teams by managing their other responsibilities with powerful features like incident auto-remediation and auto service-request fulfilment. This gives them the free time to develop their expertise in these new systems. In addition, there is serviceability. KubeEdege is a viable alternative to products like Microsoft Azure IoT Edge or AWS IoT Greengrass for IoT-related applications. However, enterprises can reach out to Microsoft Azure or AWS for support for issues and have a guaranteed level of service specified in their SLAs. KubeEdge, being open-source, lacks this level of serviceability. With AIOps managing and maintaining KubeEdge configurations and workflows, enterprises can recreate the level of serviceability expected from the top competitors.
Then there's the issue of security. Kubernetes has many challenges related to security - with container images and improper configurations being the chief concerns. If a compromised container image is replicated onto thousands of edge nodes with KubeEdge, the attack surface and exposure are just as exponentially higher and the risk correspondingly greater. AIOps helps here by building a Zero Trust system by monitoring entire edge nodes and cloud servers with anomaly detection. This enables it to catch unexpected behaviors in particular edge nodes and create meaningful alerts for ITOps teams.
There are many complementary elements in KubeEdge and AIOps platforms - including the focus on monitoring and agent-driven orchestration. Working together they can provide incremental benefit to enterprises that are looking to innovate how edge devices are managed dynamically. While both are powerful tools on their own, together, they can increase the value and the RoI of edge computing in an enterprise.
To learn more about Edge Cloud Management and KubeEdge Management with AIOps, please visit our Edge Cloud Observability page