notes_Tutti-acm22.pdf

Untitled

Untitled

  1. UE experiences multi-rounds of radio resource request grant operations

    1. UE proactively sends measurement reports to BS for monitoring the wireless channel quality — when new data packet arrives at UE buffer, triggers a buffer status report (BSR) message to the serving BS in step 2.
    2. BS returns a UL (Uplink) Grant message carrying the physical resource block (PRB) configuration according to the CQI and buffer status from step3 to step 9
    3. Implication:
      1. Figure 3 shows that there is belated and insufficient UL resource grants
      2. 50 BSR requests but only 12 cases are successfully granted on time
      3. Even if UE gets the UL grants, the allocated resources are not enough to clear the buffer data at once

  2. Most of the delay is due to multi-round interactions (And getting UL Grant)

    Problem Nowadays: MEC (Mobile Edge Computing) is important for 5G ecosystems as it allows for supporting demanding applications with latency requirements. (by using edge servers instead of remote cloud)

Tutti:

  1. Couples 5G RAN and MEC to satisfy the latency-critical applications
  2. Integrate application-layer informagtion on edge servers and physical link dynamics of 5G RAN

⇒ Tutti customizes service demand of each frame and proactively satisfies it by using incentive/acceleration model;

Methodology / Goal:

  1. Content-aware regression learning algorithm → Predict next frame size to couple application-layer information with the edge server
    1. Use ML (ResNET) to predict next size
  2. Predict network conditions to couple edge-server with 5G RAN
    1. Use Gaussian distribution and statistical approach (PDF, CDF)
  3. Use both information to create a demand rule

Untitled

⇒ assuming all packets are sent at once (no retransmission: r = 0)

Untitled

  1. Use incentive and acceleration model
    1. use deadline timer to judge whether current TTI is close to deadline