MirageOS Unikernel with Network Acceleration for IoT Cloud Environments
Takayuki Imada · 2018
Recently, data analysis for IoT has been done on cloud computing platforms via the Internet in many fields. Such a platform and service applications running on it need to i) be robust against any attacks as possible and ii) have higher network performance to handle increasing IoT data. We propose a scheme to accelerate the network performance of MirageOS Unikernel running on ukvm with features such as employing a type-safe functional language to reduce software vulnerability risks. Through performance evaluations of our proposal, we found that our proposal can achieve up to 2.79x and 1.97x higher network throughput scores for UDP and TCP respectively compared with the original MirageOS Unikernel on ukvm configuration. Another evaluation indicated a MQTT broker application implemented in our proposal can reduce about up to 60% of the total CPU utilization compared with an existing MQTT broker application written in C. Therefore, we confirmed that our proposal enables us to build IoT service applications running on cloud platforms with higher robustness in software layers and higher network performance.