This paper is based on rowhammer attack to DRAM chips. It revealed that reducing wordline voltage can reduce RowHammer vulnerability without significantly affecting reliable DRAM operation.
Strengths
- It is regirous to prove the effect on reducing the voltage, while not reducing basic function (DRAM operations).
- The experiment setup is very comprehensive. It almost considered all the details.
Weaknesses
- Lacks for novelty. The workload of this research and its real-world value are undeniable, however rigorous emprical experiment sometimes seems wordy and boring.
- How to decide the number of DRAM chips used in experiments. I might doubt the robustness of the experiment samples.
Can you do better?
- The main side effect of reducing wordline voltage includes worsen access latency and rentention time. There should be some trade-off between vulnerabilities and performance.
- Though it is not easy to talk many factors together, I still think we need to combine voltage reduction with other rowhammer mitigate strategies.
Takeaways
This work focus on a small topic and conducted solid experiments. It is a great example.
I really appreciate those observations. Though they might seem a little wordy.
Other Comments
I am actually qutie confused why so many manufacturers don't reduce the wordline voltage since it should be a naturlly optimal strategy? In other words, is there any other real world factors that is not ignored in this paper?