Automatic failure diagnosis for Wireless LAN systems based on distribution of responses from STAs

Abstract

With increase in the use of Wireless LANs in the enterprise area, network systems are often scattered across multiple locations in a wide area. Hence, there may be cases where a network administrator manages multiple locations. When users complain about connectivity issues, network administrators are expected to diagnose these problems without disrupting the normal operation of devices such as wireless stations (STAs) or access points (APs). In such cases, an on-site failure diagnosis is both time-consuming and expensive. Thus, an automated diagnostic technique to identify the cause of failure is important. We propose a method that can automatically identify the cause of the connectivity issues. The proposed method does not introduce any additional monitoring device, or require any functionality to be added in the AP or STA. In our proposed method, we send test packets to all STAs periodically and analyze the distribution of the responses and the RSSIs (received signal strength indicators) of the STAs. This helps to distinguish whether the connectivity issue is being caused by an AP failure, STA failure, or by a coverage-hole, which is a lack of network connectivity in certain areas of a building. The experimental results show that our proposed method is effective for the detection of AP failures, STA failures, and coverage-hole failures. From the detection results of a long experimental trial conducted in an actual office environment, it was observed that the false-detection probability of AP failure was very low. The STA failure was reported to be 0.36%, which was assumed to be because of hardware switch-offs or false detections because STA breakdowns were not reported during the trial. The ratio of the status “AP failure or STA failure” in the total failure detection was about 11%. Therefore, our proposed diagnosis system can suggest the cause of failure with a probability of 89%.