Research

Investigating SSD aging process and it’s influence on performance

  • Ongoing
  • Sep, 2024
  • TBD

SSDs are designed to wear out from usage. As SSDs get older, they will have degradation in performance such as smaller bandwidth and larger latency. Although some research has noticed performance degradation in old SSDs, we have’t seen much research going deeper to investigate problems such as ‘Which performance metrics are influenced by aging, and in what ways are they affected?’, ‘What factors will affect the SSD aging process and how will they affect it?’, ‘Is there a framework to reason about an SSD’s lifetime?’ Our work aims to investigate those questions to provide a better understanding about SSD aging process and help utilize SSDs for a longer time. This project is advised by Timmy and Anshul.

Sustainability

Performance Degradation

SSD

Automatic fine-grained instrumentation of distributed systems tracing for performance diagnosis

  • Ongoing

This project, which is led by my labmate Julia Cusatis, aims to automatically detect performance issues in distributed systems. The tool AutoTrace instruments and compiles code to get runtime traces, and then it analyze trace data to find out slow parts in the code.

Performance Debugging

Code Instrumentation

Distributed System Tracing

Synthesizing Runtime Checkers from Tests to Detect Semantic Failures in Production

  • Finished

Silent semantic failure detection is a challenging topic in system reliability. While other types of semantic failures, such as fail-stop and fail-slow failures, are relatively easy to detect by checking heartbeats and timeouts, silent semantic failures violate semantics but remain operational without explicit error signals, making detection a challenge without a deep understanding of system semantics. To address this issue, this project aimed to synthesize runtime checkers with minimal manual effort by deriving them automatically from existing tests in the source code. I work in a five-member group leaded by Chang Lou and supervised by Ryan Huang.

Distributed System Reliability

Semantic Failures

Code Instrumentation