- This topic has 4 replies, 1 voice, and was last updated 9 years, 8 months ago by ALU.
10th May 2011 at 18:45 #66339ALUGuest
I am suspecting that the Alcatel indicator for throughput (GPRS/EGPRS) is wrong. I conducted a test by forcing the MS on GPRS and downloading data continuously for a few hours.On TEMS the average throughput that I got was around 65-70Kbps(with Max being 80Kbps for 4TSs used).However, for the same interval the throughput per TBF “GPRS_DL_Useful_Throughput_Radio_GPRS_TBF_Avg” that was observed on NPO was around 250Kbps which is strange as in GPRS the max throughput that you can achieve would be around 80Kbps( CS4 and 4 TSs being used). Would you please share your views in this regard and has anyone observed any discrepancy in Throughput formula?10th May 2011 at 19:23 #66340pixGuest
in which release are you ?
this indicator’s formula is based on a total amount of volume, divided by a time. Check those two values, see which one is “weird”. Maybe you’ll be able to investigate where is the fault coming from.
since B10, you can check the LLC throughput distribution (in the report mono object distribution), which is probably more precise.
i don’t recall any issue with the average throughput per tbf though… but i never checked it as you did 🙂
pix10th May 2011 at 21:32 #66341ALUGuest
Thanx Pix for your quick response.We have B-10 in the network. I have already checked the counter level stats and I see high values for data volume(bytes) for CS4 resulting in high Throughput.But how can we be sure that the value reported for data transferred is correct or not? Secondly, I seriously doubt the reliability of Alcatel Throughput Formula for both GPRS/EGPRS(per TBF and even per cell). In our network, normally we get very good throughput from Drive Test(above 100Kbps) but no BSC is exceeding 40Kbps EGPRS Throughput per TBF which is very strange. Note that as per our investigations, same behaviour is being observed in other Alcatel Networks. I can confirm that there is no limitation from any interface point of view.We have already done expansions for Abis,PS Atermux,Gb interfaces and tried almost all Radio Features aswell but to no avail.So,authenticity of this indicator is quite doubtful(I personally believe that in the formula,there is some problem with the denominator(observation period)!11th May 2011 at 15:31 #66342pixGuest
the denominator is doubtful too, in my opinion 🙂
GPRS_DL_useful_throughput_radio_GPRS_TBF_avg = GPRS_DL_useful_bits_CSx_ack / (GPRS_DL_active_connection_GPRS_ack_time*1000)
GPRS_DL_active_connection_GPRS_ack_time = Cumulated time duration of all active DL TBFs established in GPRS mode and RLC acknowledged mode
Note: An active DL TBF connection is a DL TBF not in delayed release state.
Are all your TBF in RLC ACK mode ? As far as I can see, that would be the only possible explanation.
Have you checked the mono object distribution reports I was talking about ?
Basically, you could also check these (as new, but simpler) indicators:
They should provide better stats.
let me know…
pix14th May 2011 at 08:44 #66343ALUGuest
Yes, all the TBFs are in Acknowledged mode.
Interestingly, the LLC Throughput indicators “GPRS_DL_LLC_throughput_per_GPRS/EGPRS_ack_TBF” give even stranger values. The values for the BSCs done exceed even 5Kbps :). Also, note that the denominator used in this formula is the same one that we are suspecting to be wrong (GPRS_DL_active_connection_EGPRS_ack_time*1000). So, we cannot expect even this indicator to give correct values. I wonder if someone has devised a new formula with new observation period which depicts the true throughput values?