Hints for Memory Optimization of Asylo-based SGX Enclave Applications

Abstract

Intel SGX is a technology for providing the hardware-based trusted execution environment, that is available on the 6th generation Intel CPUs and above. Despite the strong security model, SGX has a fundamental limitation in its memory capacity. An SGX-enabled processor has 128 MB of the trusted memory region, called enclave page cache. Due to the limited EPC size, loading and running memory-intensive secure applications is limited. We can use SGX paging, adjusting heap sizes to accommodate secure applications in SGX enclaves. However, the performance impact of memory pressure has not been thoroughly investigated. In this paper, we analyze the relationship between the memory footprint of the secure application with the performance. According to our analysis, we show that the memory footprint of the enclave heap could negatively affect the performance of the latency-sensitive application. Besides, we show that we can effectively mitigate performance degradation by adjusting the maximum heap size.