questions about latency of small request size and large request size
Opened this issue · 0 comments
Hi, I'm writing this because I'm just wondering what do you think about my problem which is related to MQSim. Let me get to the point directly.
As shown in your FAST18 paper's figure 10 and the experimental results I made from MQSim, the latency (I mean, device response time in MQSim's result XML file) of small size (e.g., 8kB) is smaller than that of a large request size (e.g., 512KB).
However, figure 3 of the paper from NVMOS'20 [1] showed that the latency of 4KB is much larger than that of 128KB on both ZNS and traditional SSD. And they said that this is because of SSD's internal parallelism.
It seems that this paper shows the opposite results from yours.
I took a look at your source code and it looks like your code also considers the internal parallelism of SSD. But I don't know why they have different results, I mean totally opposite.
So I just wonder what do you think about this.
Is there anything I understood wrong? if so, please let me know.
[1] Exploring Performance Characteristics of ZNS SSDs: Observation and implication.