search

bobzhuyb / ns3-rdma

NS3 simulator for RDMA over Converged Ethernet v2 (RoCEv2), including the implementation of DCQCN, TIMELY, PFC, ECN and shared buffer switch
GNU General Public License v2.0
260 stars 119 forks source link

a specific derivation process for R_AI #40

Open BYyzt opened 1 year ago

BYyzt commented 1 year ago

Hello, I would like to inquire about the rate increment R_AI during the active increase process of the DCQCN algorithm.Is there a specific derivation process for R_AI , and what factors are related to the determination of this rate increment? Is there a specific expression? It doesn't seem to be mentioned in the article.

bobzhuyb commented 1 year ago

R_AI a trade-off between convergence speed and the oscillation around the convergence point. Oscillation sometimes impacts utilization. It's often tuned empirically.

BYyzt commented 1 year ago

I would like to try to establish some connection between rate increment R_AI and queue congestion information, that rate increment R_AI becomes a function of queue congestion information. Can you provide some suggestions and guidance?Thanks.

BYyzt commented 1 year ago

Hello, I am a bit confused about the rate change in the code. For example, how is the 8 in the following code determined? Is there a specific derivation? Looking forward to your reply

Void QbbNetDevice:: AdjustRates (uint32_t fIndex, uint32_t hop, DataRate increase)

{If ((m rpByteStage [fIndex] [hop]=1) | (m rpTimeStage [fIndex] [hop]=1))&&(m targetRate [fIndex] [hop]>10 * m rateAll [fIndex] [hop]) M_ TargetRate [fIndex] [hop]/=8;