Created attachment 2976 [details]
test script

Hi, 

In a downlink scenario CqaFfMacScheduler scheduler is assigning all the RBGs to a single user even though there is little data to be transmitted to it. In the end there is only one UE scheduled during one TTI.

In the attached test script I am generating 200 byte packet, so in total,

200 (payload) + 2 (RLC header) + 2 (PDCP header) + 20 (IP header) + 8 (UDP header) = 232 bytes or 1856 bits.

If the MCS is 28, the maximum TB size for 24 PRBs is 2196 bytes but to TX 232 bytes a TB of size 373 bytes would be enough. With this bug in MAC stats we can see that every UE has been assigned a maximum TB size, e.g.,

% time    cellId    IMSI    frame    sframe    RNTI    mcsTb1    sizeTb1   
 0.3        1        2        31       1       1        28        2196            
0.301       1        3        31       2       2        28        2196            
0.302       1        1        31       3       3        28        2196

The reason for this behaviour is incorrect comparison of needed resources and assigned resources at line 1488

if (UeToAmountOfDataToTransfer.find (userWithMaximumMetric)->second <= UeToAmountOfAssignedResources.find (userWithMaximumMetric)->second*tolerance)
//||(UeHasReachedGBR.find(userWithMaximumMetric)->second == true))
  {
    itCurrentGroup->second.erase (userWithMaximumMetric);
  }

Here UeToAmountOfDataToTransfer is in "bits" while "UeToAmountOfAssignedResources" is in bytes. Due to this reason a selected UE who already has been assigned sufficient amount of RBGs never get erased and the while loop "while (availableRBGs.size ()>0 and itCurrentGroup->second.size ()>0)" at line 1310, kept running till all the RBGs are assigned to a single UE.  

Kind regards,
Zoraze