Question On Quality of Service - 3G/EDGE
Hi people, I have a question on the word and actual meaning of quality of service.. This is the scenario: == The customer here bought a 3.6 mbps link from the ISP to provide 3G and EDGE services for mobile users. And we have an online tool : http://www.speedtest.net/ Which tests your active internet link - in this case we put the 3G/HSDPA cards [from Huawei] that are so popular in Kenya at the moment - and the dial on the webpage moves to a certain high. However, we cannot get more than 2.7-2.9 mbps for some sites linked with an STM-1 link [ a site here refers to a base station or radio base station for WCDMA] And cannot get more than 1.6 mbps for a radio base station linked with 2 E1s. On the GPRS/WCDMA equipment - I have set quality of service [QOS] for HSDPA to be 2048kbps [2mpbs] Uplink ---- and 4096kbps [4mbps] downlink which will cater nicely for their 3.6 mbps dedicated link to the ISP. === Now the customer is convinced that the customers with 3G cards should get more than 2mbps for the sites with 2 E1 links --- but we cannot get higher than 1.6 mbps. Mind you we cannot even get more than 2.8 mpbs for the sites with an STM-1 link... but customer thinks we should get the same as what the ISP sends here [3.6 mbps] per mobile or 3G/HSDPA card user. Now I cannot pin-point the exact theory and calculations so that I can offer the explanation as to why customers cannot get the full 3.6mbps from ISP.... I am also not sure of the mathematics to explain why the 2E1s sites are getting lower than the sites with a full STM-1 link - : but I do know that 1E1 with 31 [timeslots or channels] = 64kbps X 31 ---> one for upload and the other download - Giving you a maximum of 1.98 mbps - but we are getting a high 1.6 mbps in a 3G/HSDPA card for these sites with no STM-1 links. ==== Anybody care to shed some light on these calculations from ISP to end users ? And what an end user will get ? I am having alot of grey areas and assumptions which are not very satisfactory.
Hi Stephen, First off....The technology is a shared one and so there's always a chance that the bandwidth would be intermittent depending on the population density... The physical links to the site serve both the user plane(actual data) and the control plane(signalling)....Depending on how the operator has configured, the allocation of timeslots for data and voice vary(in case of EDGE)..voice will always take priority over data and this means voice will be pre-allocated more timeslots than data.... QoS can be set at the radio or core network....At the radio network, channels will be set according to the number of cells per site....In high density areas, the cells are split into micro-cells to cater for the population whereas the opposite is true for rural area....So consider that the 2E1s per site will be shared among e.g. 3 cells. If the 2 E1s are fully set for data then that means theroritically 2 *2Mbps = 4Mbps.... So you can already see that guaranteeing 3.6Mbps is already a problem here!!! QoS can also be set at the core network...The HLRs subscriber profile can be set to allow 512kbps/1Mbps/2Mbps..etc per subscriber....Such that during data call setup, the SGSN will get the user subscription from the HLR and negotiate that QoS with the RNC/BSC and/or NodeB/BTS.... To my knowledge, most mobile operators here(Saf.Zain/Orange/Yu) hardly use QoS but allow for "best-effort" contention among the users....So your modem may support 3.6Mbps max but there are all sorts of bottlenecks(a.k.a policies) to contend with.... ./bernard On Wed, Aug 19, 2009 at 11:49 AM, ndungu stephen <ndungustephen@gmail.com>wrote:
Hi people,
I have a question on the word and actual meaning of quality of service..
This is the scenario: ==
The customer here bought a 3.6 mbps link from the ISP to provide 3G and EDGE services for mobile users.
And we have an online tool :
Which tests your active internet link - in this case we put the 3G/HSDPA cards [from Huawei] that are so popular in Kenya at the moment - and the dial on the webpage moves to a certain high.
However, we cannot get more than 2.7-2.9 mbps for some sites linked with an STM-1 link [ a site here refers to a base station or radio base station for WCDMA] And cannot get more than 1.6 mbps for a radio base station linked with 2 E1s.
On the GPRS/WCDMA equipment - I have set quality of service [QOS] for HSDPA to be 2048kbps [2mpbs] Uplink ---- and 4096kbps [4mbps] downlink which will cater nicely for their 3.6 mbps dedicated link to the ISP.
=== Now the customer is convinced that the customers with 3G cards should get more than 2mbps for the sites with 2 E1 links --- but we cannot get higher than 1.6 mbps.
Mind you we cannot even get more than 2.8 mpbs for the sites with an STM-1 link... but customer thinks we should get the same as what the ISP sends here [3.6 mbps] per mobile or 3G/HSDPA card user.
Now I cannot pin-point the exact theory and calculations so that I can offer the explanation as to why customers cannot get the full 3.6mbps from ISP....
I am also not sure of the mathematics to explain why the 2E1s sites are getting lower than the sites with a full STM-1 link -
: but I do know that 1E1 with 31 [timeslots or channels] = 64kbps X 31 ---> one for upload and the other download -
Giving you a maximum of 1.98 mbps - but we are getting a high 1.6 mbps in a 3G/HSDPA card for these sites with no STM-1 links.
====
Anybody care to shed some light on these calculations from ISP to end users ? And what an end user will get ?
I am having alot of grey areas and assumptions which are not very satisfactory.
_______________________________________________ Skunkworks mailing list Skunkworks@lists.my.co.ke http://lists.my.co.ke/cgi-bin/mailman/listinfo/skunkworks Other services @ http://my.co.ke Other lists ------------- Announce: http://lists.my.co.ke/cgi-bin/mailman/listinfo/skunkworks-announce Science: http://lists.my.co.ke/cgi-bin/mailman/listinfo/science kazi: http://lists.my.co.ke/cgi-bin/mailman/admin/kazi/general
Thanks to all who sent their answers -- the calculations and arguments are now settled. w/r Stephen N.
just to add though replies were off list... 1 E1 is 31 time slots yes but most of the times only 30 is used for payload max, atleast 1 is used for signaling this is incase of RBS/BTSs .... 2009/8/19 ndungu stephen <ndungustephen@gmail.com>
Thanks to all who sent their answers -- the calculations and arguments are now settled.
w/r Stephen N.
_______________________________________________ Skunkworks mailing list Skunkworks@lists.my.co.ke http://lists.my.co.ke/cgi-bin/mailman/listinfo/skunkworks Other services @ http://my.co.ke Other lists ------------- Announce: http://lists.my.co.ke/cgi-bin/mailman/listinfo/skunkworks-announce Science: http://lists.my.co.ke/cgi-bin/mailman/listinfo/science kazi: http://lists.my.co.ke/cgi-bin/mailman/admin/kazi/general
-- Watson wanjohi kambo
@Ndungu, I think the problem lies in the modulation and coding setup of the mux. Possibly check out the differences between the two different vendors modulation schemes, if they are different platforms. HTH's .
On Wed, Aug 19, 2009 at 12:39 PM, aki <aki275@googlemail.com> wrote:
@Ndungu, I think the problem lies in the modulation and coding setup of the mux. Possibly check out the differences between the two different vendors modulation schemes, if they are different platforms. HTH's .
@Ndungu further to the above, does the Mux give zones reports on load data, saturation etc? I think you have 2 issues. One, as I mention above as Mux issue or that sectors of radio are loaded thus less throughputs. Best way to handle such a scenario would be to perform analysis on the Mux device. Get those stats and base some calculations on STM1 to load distribution. If you have the cell data, then building a QOS based on traffic analysis, not just bandwidth will be much easier. Have I ever setup one on GSM, none ( you are quite fortunate to have access to such ). However, from old experience on some muxes in microwaves, these were some issues that I came across. And dvb-ips with return via sat had similar issues. Distance is also a major problem. Question : Is it possible to do a loop test from the Mux, radio and Modem? If yes, I suggest you start here with bench marking. HTHs. -- Sign on my car is " L " plates : c#.net
Thanks for the information Bernard and aki - it gives me a place to start... offlist -->
But what I was asking is - are my assumptions correct ? The call flow looks like this [I have skipped some equipment] : ? 2E1 3.6 mbps MS--------->RBS--------------->radio link----------------->GPRS------------------->Internet ? 2E1 3.6 mbps MS<---------RBS<--------------radio link<----------------GPRS<-------------------Internet So, when calculating bandwidth for an end user - several possible calculations arise - assuming the RBS has 3 cells 1) take the two E1s to mean that - 1 is for download, the other is for upload [when there is no congestion] End user will receive = 64kbps * 31 timeslots = 1.984 mbps [not explicitly cos 1mb=1024] But end user is getting a high of 1.6 mbps here - why ? 2) take the two E1s to be shared for both down and upload End user will receive = 64kbps * 31 timeslots * 2 = 3.968 mbps [not explicitly cos 1mb=1024] End user is getting 1.6 mbps because its same scenario or position as the above 3) take that the E1 is divided amongst the 3 cells on an RBS to serve on demand - -and if 1E1 is used for download and the other for upload. So assuming one cell has no user: a) 2/3 * 64 kbps * 31 timeslots = 1.3 mbps or if each cells have at least 1 user b) 1/3 * 64 kbps * 31 timeslots = 661 kbps 4) take the two E1s to be shared for both down and upload and is divided for 3 cells of the RBS So assuming one cell has no user: a) 2/3 * 64 kbps * 31 timeslots * 2 = 2.6 mbps or if each cells have at least 1 user b) 1/3 * 64 kbps * 31 timeslots * 2 = 1.32 mbps ===
From the above results - the results in red are an approximate of what we are getting on the ground.
However, I need to come up with a comprehensive and "almost" true way to calculate the above, in-order to advise appropriately. So based on your experiences, which is the right way to calculate - out of the 4 methods ? These are also microwave links, so take it as same scenario. === How will the increase in users affect these calculations of bandwidth ? I still do not understand the concept of a "dedicated link" when it comes to user distribution - in retrospect - if the customer is getting 3.6 mbps from ISP and he is to distribute to mobile users - are the users really supposed to get 3.6 mbps each - or will this now be divided by the number of active users? Because in theory, each of these users has this 4mbps capable on his 3G/HSDPA card for radio link- and each user has a 3.6 mbps dedicated link to internet on the operator's side. so: is it: 4mbps 3.6mbps user ----------->radio------------------->intenet or is it 4mbps [3.6mbps divided by no. of users] user------------>radio------------------->internet And how does uplink and downlink separate channels to internet affect this ? === I know its a long question - sorry for the time and length. w/r Stephen N.
On Wed, Aug 19, 2009 at 3:10 PM, ndungu stephen <ndungustephen@gmail.com>wrote:
But what I was asking is - are my assumptions correct ?
The call flow looks like this [I have skipped some equipment] :
@Ndungu, thanks much for a challenging post, its quite interesting and I intend to fully dig deeper. I've replied offlist. :-)
participants (4)
-
aki -
Bernard Mwagiru -
ndungu stephen -
Watson Kambo