User Throughput versus Cell throughput

Don’t know what to say…

Other cell have also important data on last slot.
I mean high traffic on last slot.
You can check in your network too.
This means reduced traffic coming from chit-chat data like facebook messenger, whataspp, email, linkedin, etc.
Not heavy downloaders.

1 Like

We do not have 5G running here in Brazil, but for 4G this formula uses to work quite well.

I doubt it works well if user throughput is higher than cell’s throughput even if Huawei found an explanation for it. :slight_smile:

Maybe, the two formulas should be considered, in order to a have a kind of better of traffic outlook perspective.

Indeed, this could be a solution, network wide to compare which one is closer to reality.
In last slot is almost 50% of volume of traffic, how can we remove it from formula even if 3GPP recommends this?

I am posting again the graph:

It looks much better user throughput compared to cell throughput for same cell:

Cell has no capacity issue as MU-MIMO is active with 8 layers.

As one can not user throughput is somehow constant in time which is normal it depends on RF conditions of users.
If users are static throughput will be the same,more or less.

Just to add every vendor having different algorithm for throughput calculation.
Once we done comparison and seen Huawei no 1.
Nokia 2 Mbps for cell where huawei showing 9 Mbps.

So I think I will keep it without last slot, definitely.

Ericsson just scheduled users in one go.
Effectively less time.
User Throughput inversely proportional to time.
So main focus is that how vendor algorithm can shorten time.

I am still not able to find counter for number of slots that scheduler sent data in DL.
Having that one it will be easy to calculate users scheduled per slot and much more.
But where to get that counter from?

For those who are interested I have found the 5G counter of Huawei which is giving the number of slots sched in DL:
N.ThpTime.DL.Cell

Accordingly number of users sched per tti is given by this formula:
N.ThpTime.DL / N.ThpTime.DL.Cell

Also percentage of scheduled slots in DL for 1 hour is N.ThpTime.DL.Cell/(3600*1000000)

Last TTI is usually underutilized.
So if the % of last TTI is high, it is typical to have Cell Throughput lower than User Throughput.

Example:
A single user in the cell that utilizes 2 TTI (1000 bits in 1st TTI and 100 bits in last TTI)
Then User Throughput is 1 Mbps while Cell Throughput is 550 kbps.

Removing last TTI is recommended by 3GPP.
Unfortunately it favors cell edge since cell center user may empty buffer in one TTI (BW dependent and packet size dependent).

1 Like

Time.cell pegs ONCE for every TTI used (1 or more users).
So percentage of used slots could be Time.Cell/ measurement period.

1 Like

Guys,
Any summary for whats is basic difference between User Throughput and Cell Throughput?

Very basic difference is:

  1. User Throughput is calculated per QCI at RLC layer.
  1. Cell Throughput is overall Throughput calculated at MAC.

Apart from that, there is difference in pegging these counters.

User Throughput like vol/time per qci on which traffic there.
Cell Throughput vol/active tti.

User Throughput calculated to give user experience for Throughput, generally User Throughput on rough side = Cell Throughput / number of active users in cell.

Let me give you a raw example from counter pegging point of view:

It is importatn to note that User Throughput time starts when first user is scheduled and for Cell Throughput time starts when data arrives in cell DL buffer. So in above case User Throughput = (44 Mega bits / 30 msec ) = 1.46
And Cell Throughput = 44 / 12 = 3.66

This is raw example.
But User Throughput also depends on counter pegging, for example check this figure:

User Throughput also depends on the counters you are using.
Hope it clarifies your concerns.

No one can beat huawei in user thrpt values…:grin:
Magic is there.

Tricky and difficult to understand. :blush:

1 Like

Hi dears, if you have any query please feel free to ask.
I tried to differentiate things with colors ,so that it’s easier to understand.

Great @AbdurRehman87!

So throughput per user will be = (L.thrp.bits.dl-l.thrp.bits.dl.last TTi)/ l.thrp.bits.dl.rmv.last TTI